[go: up one dir, main page]

CN112315459B - A multifunctional digital evaluation intelligent seat - Google Patents

A multifunctional digital evaluation intelligent seat Download PDF

Info

Publication number
CN112315459B
CN112315459B CN202011345922.8A CN202011345922A CN112315459B CN 112315459 B CN112315459 B CN 112315459B CN 202011345922 A CN202011345922 A CN 202011345922A CN 112315459 B CN112315459 B CN 112315459B
Authority
CN
China
Prior art keywords
data
stage
armrest
tremor
gravity
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202011345922.8A
Other languages
Chinese (zh)
Other versions
CN112315459A (en
Inventor
林瑞峰
胡悦鹏
程敬原
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Science and Technology of China USTC
Original Assignee
University of Science and Technology of China USTC
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by University of Science and Technology of China USTC filed Critical University of Science and Technology of China USTC
Priority to CN202011345922.8A priority Critical patent/CN112315459B/en
Publication of CN112315459A publication Critical patent/CN112315459A/en
Application granted granted Critical
Publication of CN112315459B publication Critical patent/CN112315459B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/103—Measuring devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
    • A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor or mobility of a limb
    • A61B5/1101—Detecting tremor
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/40—Detecting, measuring or recording for evaluating the nervous system
    • A61B5/4076—Diagnosing or monitoring particular conditions of the nervous system
    • A61B5/4082—Diagnosing or monitoring movement diseases, e.g. Parkinson, Huntington or Tourette
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/68—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
    • A61B5/6887—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient mounted on external non-worn devices, e.g. non-medical devices
    • A61B5/6891—Furniture
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/725—Details of waveform analysis using specific filters therefor, e.g. Kalman or adaptive filters
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Veterinary Medicine (AREA)
  • Public Health (AREA)
  • Animal Behavior & Ethology (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Medical Informatics (AREA)
  • Surgery (AREA)
  • Physiology (AREA)
  • Signal Processing (AREA)
  • Artificial Intelligence (AREA)
  • Neurology (AREA)
  • Psychiatry (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Neurosurgery (AREA)
  • Evolutionary Computation (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Developmental Disabilities (AREA)
  • Dentistry (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

本发明提出一种多功能数字化评估智能座椅，包括：座椅本体、底座、坐垫部位、扶手部位；在智能座椅坐垫部位、以及扶手部位分别安装柔性压阻传感器，在扶手部位加装伺服电机编码器以及加速度传感器；所述对柔性压阻传感器用于获取表面上方接触面压力分布变化数据；其中，两层导电层呈九十度水平垂直排列，并在两层导电层中间放置一层所述压敏电阻层，随后用缝纫机将三层车线固定，并使用热熔胶进行加固；所述伺服电机编码用于获取手臂旋转数据，扶手部位加速度传感器用于测量手臂的震动数据；上述各数据通过有线串口协议传输，由下位机发送至上位机，并在上位机对数据进行处理分析，得到被采集病患的相关参数，并将数据与分析结果进行存储。

Figure 202011345922

The invention provides a multifunctional digital evaluation intelligent seat, which includes: a seat body, a base, a seat cushion part, and an armrest part; flexible piezoresistive sensors are respectively installed on the seat cushion part and the armrest part of the intelligent seat, and a servo is installed at the armrest part. A motor encoder and an acceleration sensor; the pair of flexible piezoresistive sensors is used to obtain the change data of the contact surface pressure distribution above the surface; wherein, the two conductive layers are arranged horizontally and vertically at 90 degrees, and a layer is placed in the middle of the two conductive layers The piezoresistor layer is then fixed with a sewing machine for three layers of car lines and reinforced with hot melt adhesive; the servo motor code is used to obtain the arm rotation data, and the acceleration sensor at the armrest is used to measure the arm vibration data; the above-mentioned Each data is transmitted through the wired serial port protocol, and sent from the lower computer to the upper computer, and the data is processed and analyzed in the upper computer to obtain the relevant parameters of the collected patients, and the data and analysis results are stored.

Figure 202011345922

Description

Multifunctional digital assessment intelligent seat
Technical Field
The invention belongs to the field of medical instruments, and particularly relates to a multifunctional digital evaluation intelligent seat.
Background
The assessment of the dyskinesia patients is mostly diagnosed and assessed by clinical medical staff through subjective experience, for example, the assessment and diagnosis method of the Parkinson disease patients comprises the following steps: according to assessment items and standards in a unified Parkinson disease assessment scale (UPDRS) published by the International dyskinesia Association (MDS), medical staff are obtained by experience observation and judgment, namely, indexes such as tremor degree, muscle rigidity degree and unbalance degree of bilateral limbs of Parkinson disease patients are observed and assessed through naked eyes so as to carry out subsequent diagnosis, subjective factors exist, and the tremor amplitude, the muscle rigidity and the unbalance degree cannot be accurately quantified, and the defects exist. The existing assessment device mainly performs tremor assessment through an acceleration sensor, but has the defects of inconvenient wearing and capability of only assessing a single symptom.
Disclosure of Invention
In order to solve the technical problems, the flexible piezoresistive sensors, the servo motor and the inertial sensor are arranged on the seat cushion and the armrest parts of the special seat, and partial symptoms of the dyskinesia disease are evaluated, including but not limited to measuring the tremor degree, the muscle stiffness degree, the limb discordance degree, the rising force condition and the like of a patient, so that the aim of objectively and accurately evaluating the illness state of the dyskinesia disease patient including the Parkinson patient is fulfilled, and the special seat has the advantages of reproducibility, storage and analysis. The invention installs flexible piezoresistance sensors on the seat cushion and the armrest part of the special seat, installs a motor encoder and an acceleration sensor on the armrest part, transmits pressure distribution change data of a contact surface above the sensors, motor rotation data and armrest part inertial sensor data to an upper computer through wired serial port protocol transmission, processes and analyzes the data by the upper computer to obtain relevant parameters of a collected patient, and stores the data and an analysis result.
The technical scheme of the invention is that the multifunctional digital evaluation intelligent seat comprises: the seat comprises a seat body, a base, a cushion part and an armrest part;
the intelligent seat is characterized in that flexible piezoresistive sensors are respectively arranged at the seat cushion part and the armrest part of the intelligent seat, and a servo motor encoder and an acceleration sensor are additionally arranged at the armrest part;
the flexible piezoresistive sensor is used for acquiring pressure distribution change data of a contact surface above the surface; the flexible piezoresistive sensor comprises: the conductive layer is provided with stainless steel conductive fibers and polyethylene fibers; the two conductive layers are vertically arranged at ninety degrees horizontally, and a flexible semiconductor piezoresistor layer is arranged between the two conductive layers and is formed by staggering silicon carbide fibers and polyethylene fibers; fixing the three layers of threads by using a sewing machine, and reinforcing by using hot melt adhesive;
the servo motor codes are used for acquiring arm rotation data, and the armrest part acceleration sensor is used for measuring vibration data of the arm;
the upper computer comprises a memory and a singlechip processor;
the data are transmitted to an upper computer by a lower computer through wired transmission, the data are processed and analyzed by a single chip processor on the upper computer to obtain relevant parameters of the collected patients, and the data and the analysis result are stored and displayed.
Furthermore, the flexible piezoresistive sensor is manufactured by weaving stainless steel fibers into metal yarns, weaving the metal yarns and common yarns, namely polyethylene yarns into a plurality of metal yarn stripes in a staggered manner to form conductive layers, vertically stacking the two conductive layers to form matrix grid arrangement, placing a flexible semiconductor piezoresistor layer in the middle of the conductive layers, connecting the stripe metal bands of the matrix grid through metal snap fasteners, and connecting the stripe metal bands with external flat cables.
Furthermore, the intelligent seat comprises a base part consisting of a stainless steel disc base and a booster oil pump hydraulic lifting rod, the seat body part comprises a component consisting of a reinforced composite straight plate and high-density sponge, and the outer surface of the seat body part is covered with a wear-resistant leather material, and the stainless steel disc base is a circular rotary hydraulic base; the circular rotary hydraulic base is connected with a seat gasket through spot welding, a reinforced composite straight plate is adhered with high-density sponge which is cut flatly to form various components, the components comprise handrails, a backrest and a cushion component, the components are fixedly connected through steel nails, and finally wear-resistant leather materials are covered on the surface of the components for packaging.
Furthermore, two straight plates are respectively adopted at the left armrest and the right armrest, a servo motor encoder is fixed at the fixed ends of the straight plates, then an acceleration sensor is placed at the suspended end of the straight plates, one end of the servo motor encoder is fixed at the corresponding position of the armrest by a hinge, a flexible piezoresistive sensor is laid and fixed on the upper surface of the straight plates to form an armrest component, the flexible piezoresistive sensor is installed at the position of a cushion, and the left armrest component and the right armrest component are symmetrical.
Furthermore, a flexible piezoresistive sensor with multiple channels (such as 16 x 16) is arranged between the high-density sponge and the abrasion-resistant leather material interlayer at the armrest parts of the chairs on two sides, and is reinforced by using an adhesive to prevent the flexible piezoresistive sensor from moving; the flexible piezoresistive sensors with multiple channels (such as 32 x 32) are arranged between the wear-resistant leather material and the high-density sponge interlayer at the position of the chair cushion, and are reinforced by using an adhesive, so that no wrinkling and no deviation are ensured, and the pressure distribution change data quality is ensured.
And the lower computer driving circuit board is used for acquiring data of the flexible piezoresistive sensor, comprises a power module drive and two multiplexers, divides the voltage of the flexible piezoresistive sensor single channel by using a divider resistor to obtain an analog voltage signal, performs digital-analog signal conversion on the analog voltage signal by using a DSP (digital signal processor), finally packages the digital voltage signals of the nodes into frames, and transmits the frames to the upper computer through a wired serial port.
Furthermore, the motor encoder driver adopts a servo motor driver with proper power (approximately equal to 200W) and low voltage (approximately 36V) and high current (approximately 7A), the driver and the upper computer adopt wired transmission, the transmission content comprises rotating speed and torque, the acceleration sensor adopts an acceleration sensor, and the acceleration sensor is directly connected with the upper computer through a flat cable to transmit attitude data of two handrail parts.
Further, a memory in the upper computer stores data acquisition and processing instructions, and when executed by the processor of the singlechip, the method executes the following steps to acquire and process the data of the Parkinson's patient:
step 1, acquiring and processing tremor data;
step 2, acquiring and processing muscle stiffness data;
step 3, acquiring and processing data in the rising stage;
and 4, acquiring and processing data on both sides of the body.
Further, the step 1, tremor data acquisition and processing specifically include:
step 1.1: collecting data received by the flexible resistance sensors at the handrail parts at the two sides; firstly, a median filter is used for eliminating noise, an interpolation method is used for up-sampling, and then a Gaussian filter is used for smoothing to obtain preprocessed data;
step 1.2: inputting the preprocessed gray low-resolution video data into a convolutional neural network for feature dimensionality reduction extraction, inputting the extracted feature data into a GRU (generalized regression) cyclic neural network for classification, wherein a cross entropy loss function is used as a loss function:
Figure BDA0002799924990000031
Figure BDA0002799924990000032
where N is the number of samples, M is the number of classes, yicIs an indicator variable; p is a radical oficTo predict the probability that a sample belongs to class c, the tremor factor S is output after predictionT；
Tremor factor STThe calculation process is as above, wherein pcFor the prediction probability of the model for level c, c is the weight corresponding to that level, and the weight of the lower level can be adjusted slightly higher. (for example, 0 grade in the original tremor grade corresponds to 1.2, 1 corresponds to 2.2, 2 corresponds to 2.1, 3 corresponds to 3, and 4 corresponds to 4, so that the aim of improving the conservative evaluation of medical staff on tremor invisible to the naked eye and improving the overall evaluation performance of equipment is achieved) the tremor factor has a strong relationship with the tremor performance of an evaluation object, and the tremor amplitude and frequency of the evaluation object can be better reflected.
The model training data can be from the marking data of the professional and can also be from the predefined data, and according to the severity of tremor of the evaluation object, the invention can divide the tremor into a plurality of grades and label the grades, so that the tremor can be evaluated more comprehensively, as shown in fig. 5.
Further, the tremor data acquisition and processing is optional, and another method is adopted, which comprises the following steps:
step 1.1: collecting data received by the flexible resistance sensors at the handrail parts at the two sides; firstly, filtering medium-high frequency noise in data by using a bilateral filter to obtain preprocessed data;
step 1.2: inputting each frame into a self-encoder to extract and reduce the dimension of the frame characteristics for the preprocessed gray low-resolution video data, and splicing the extracted characteristic data of each frame according to the sequence of the frame numbers to form a data matrix containing all the data characteristics; inputting the matrix into a support vector machine, judging whether a prediction error occurs or not if the data matrix is used for predicting the tremor degree, if the prediction error occurs, outputting 0 by the support vector machine, and if the prediction error does not occur, outputting 1 by the support vector machine; if the output of the data matrix obtained in the support vector machine is 1, inputting the data matrix into a convolutional neural network for time sequence feature extraction and classification, wherein the loss function uses a multi-class cross entropy function:
Figure BDA0002799924990000041
yicis an indicator variable (1 if the class sample is the same as the class of sample i, otherwise 0); p is a radical oficIs the prediction probability for a prediction sample belonging to class c. If the output of the data matrix obtained in the support vector machine is 0, the data of the batch is not classified, and the next batch of data is waited. After classifying the data, output the tremor factor:
Figure BDA0002799924990000042
wherein P isSVMFor support vector machine classification accuracy on training set, lkFor the prediction of tremor extent of this batch of data,/iDegree of tremor corresponding to category i, piM is the set of all tremor degrees, which is the predicted probability that the sample belongs to category i.
In the data processing model, the bilateral filter considers the intensity difference and the spatial relationship between data points at the same time, so that the bilateral filter can effectively filter out high-frequency and medium-frequency noise, and reserve the edge of the effective part of input data to avoid the ringing effect as much as possible. The self-encoder can extract the main features of the input data to the maximum extent, reduce the dimensionality of the input data and save computing resources for subsequent processing. The support vector machine plays the role of a data filter in this model. The method can filter the data which can be classified wrongly in the subsequent convolutional neural network model, and only the data which can be classified correctly can pass through, so that the accuracy of the model prediction can be improved. Convolutional layers in convolutional neural networks are used to extract timing features in the data, and fully-connected layers classify the data according to the timing features.
Further, the step 2, the acquisition and processing of the muscle stiffness data specifically includes:
step 2.1, collecting data received by the flexible resistance sensors at the handrail parts at the two sides; firstly, carrying out the same data preprocessing in the step 1.1;
step 2.2: summing the pressures obtained at the flexible resistance sensors at the positions of the single-side handrails, namely obtaining the total pressure at the positions of the handrails through the data of 16 channels by 16 channels; linear regression is performed by the encoder angle and the modulus of the acceleration sensor in three linear directions:
Figure BDA0002799924990000051
Figure BDA0002799924990000052
wherein R is the rigidity grade in the UPDRS scale;
when outputting, removing R0, and outputting as long as the latter part is used as a characteristic value as a stiffness factor:
Figure BDA0002799924990000053
the output value is a strong direct factor and is a multi-sensor fusion value R0Alpha, beta and lambda are regression variables, P is the total pressure value of the handrail, S is the rotation angle of the encoder,
Figure BDA0002799924990000054
is the mode of the three-dimensional accelerometer.
Further, the acquiring and processing of the data in the starting stage of step 3 specifically includes:
the rising data acquisition stage is divided into three stages: a gravity center transfer stage, a foot supporting stage and a separating rising stage; and (4) calculating the total pressure value and the rising time of each stage of the three stages respectively.
Three-stage division:
A. a gravity center transfer stage: the human body is supported by the armrests, the gravity center is gradually transferred from the buttocks to the feet, but the buttocks still have gravity. The behavior on the flexible pressure sensor data is as follows: the total pressure value of the sensor of the cushion part is reduced, the gravity center is changed, and the stage can be judged by setting a gravity center change threshold value.
B. A step supporting stage: the human body is supported by the armrests, the buttocks are separated from the chair, and no gravity exists in the cushion part. The expression on the flexible pressure sensor data is that the total pressure value of the sensor of the cushion part is zero, the total pressure sensor of the armrest part is obviously increased compared with the first stage, and the first stage and the second stage can be distinguished by judging whether the gravity of the cushion part exists or not.
C. A separation rising stage: the human body is supported by the feet, and no gravity exists in the seat cushion and the armrest part in the stage of gradually standing. The second and third stages can be distinguished by determining whether pressure is present on the armrest.
Specifically, the feature calculation at each stage is as follows:
1) total pressure value:
Figure BDA0002799924990000055
calculating the total pressure of a frame in the video-like data, wherein: i, j are pixel points, i.e. sensing points, located in i rows and j columns, m, n are the number of rows and columns of the matrix, AijPressure values of i rows and j columns;
2) change of center of gravity:
firstly, establishing a Cartesian coordinate system for a frame, adding the gravity x coordinates of all pixel points and dividing the gravity x coordinates by the number of the pixel points to obtain the gravity center position in the x direction under the condition that the pressure in the frame is greater than a specific value, obtaining the gravity center position in the y direction in the same way, and judging as a first stage if the change value of the gravity center position in the set direction exceeds a set threshold value;
3) duration of the phase: and when the gravity center change threshold frame of the first stage is reached, marking the frame as the ending frame of the first stage, recording the corresponding timestamp, and obtaining the stage duration by utilizing the two timestamps.
Further, the step 4 of collecting and processing bilateral body data specifically includes:
the tremor factor data, the muscle strength data and the maximum pressure conditions of the armrests on the two sides in the rising process of the patient are weighted and calculated, and the unbalance degree of the limbs on the two sides can be obtained:
S＝aST+bSR+cP
wherein S is a single-side limb index, a, b and c are weights of each parameter, STIs tremor factor, SRMuscle rigidity factor, P is the maximum pressure of the armrest in the rising stage; finally, the left index was used to divide the right index to obtain the degree of bilateral difference (%)
Figure BDA0002799924990000061
Has the advantages that:
1. the design scheme in the prior art is improved, and as the pressure sensors adopted in the prior art are all traditional single-point piezoresistive sensors and do not use a special driving circuit board for data acquisition, the impedance (resistance) of arm parts cannot be accurately and completely reflected, and the data refresh rate is low, the data acquisition and processing equipment for similar muscle stiffness evaluation is improved;
2. the intelligent chair carries out digital acquisition and calculation on the motion data and the rising action process of the patient, and the prior art has no similar method for calculating the type of data and has no system for combining muscle rigidity evaluation and tremor evaluation of the Parkinson patient;
3. the invention designs the special sensor, the special driving circuit board and the special seat, so that medical staff can use the device more conveniently and can evaluate the patient without sitting posture temporarily, thereby ensuring the safety of the patient to a greater degree and realizing non-invasive detection.
Drawings
FIG. 1 is a schematic view of a flexible piezoresistive sensor matrix according to the present invention;
FIG. 2 is a schematic diagram of a multifunctional digital evaluation intelligent seat according to the present invention;
FIG. 3 is a schematic view of a multifunctional digital evaluation intelligent seat structure according to the present invention;
FIG. 4 is a pressure distribution visualization display diagram;
FIG. 5 is a tremor data processing flow diagram;
FIG. 6 is a flow chart of yet another example of tremor data processing.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments, and all other embodiments obtained by a person skilled in the art based on the embodiments of the present invention belong to the protection scope of the present invention without creative efforts.
According to one embodiment of the invention, the multifunctional digital evaluation intelligent seat is provided, wherein flexible piezoresistive sensors are respectively installed at a cushion part and an armrest part of the intelligent seat, a servo motor encoder and an acceleration sensor are additionally installed at the armrest part, pressure distribution change data of a contact surface above the flexible piezoresistive sensors, rotation data of the servo motor encoder and data of the armrest part acceleration sensor are transmitted through a wired serial port protocol, the data are sent to an upper computer by a lower computer, the data are processed and analyzed by the upper computer to obtain relevant parameters of a collected patient, and the data and an analysis result are stored.
According to one embodiment of the invention, the flexible piezoresistive sensor comprises: compared with the copper-silver alloy adopted in the prior art, the stainless steel conductive fiber disclosed by the invention has the advantages that the cost can be controlled while the higher conductivity is ensured, the weight is lighter and softer, and the cost is greatly reduced; the flexible semiconductor piezoresistor layer is formed by staggering silicon carbide fibers and polyethylene fibers. Wherein, two-layer conducting layer are ninety degrees horizontal vertical arrangement to place one deck the flexible semiconductor varistor layer in two-layer conducting layer middle, can use the sewing machine to fix three-layer sewing line (guarantee that there is not the portion of puckering between three-layer intermediate layer) afterwards, and use the hot melt adhesive to consolidate. When the flexible piezoresistive sensor is manufactured, stainless steel fibers are woven into metal yarns, the metal yarns and common yarns, namely polyethylene yarns, are woven in a staggered mode to form a plurality of metal yarn stripes to form a conductive layer, two conductive layers are vertically stacked to form matrix grid arrangement, a flexible semiconductor piezoresistor layer is placed in the middle of the conductive layer, metal strips of the stripes of the matrix grid are connected with an external lead through good conductors (for example, metal snap fasteners are used as the good conductors, and connection stability can be well guaranteed), and the flexible piezoresistive sensor is manufactured;
the intelligent seat comprises a base part consisting of a stainless steel disc base and a booster oil pump hydraulic lifting rod (the lifting rod is positioned on the back of a seat and is convenient for medical staff to adjust the height), the chair body part mainly comprises a component consisting of a reinforced composite straight plate and high-density sponge, and the stainless steel disc base is a circular rotary hydraulic base covered with wear-resistant leather materials on the outer surface; the manufacturing process of the intelligent seat comprises the following steps: the circular rotary hydraulic base is connected with a seat gasket through spot welding, a reinforced composite straight plate is adhered with high-density sponge which is cut flatly to form various components, the components comprise handrails, a backrest and a cushion component, the components are fixedly connected through steel nails, and finally wear-resistant leather materials are covered on the surface of the components for packaging. The cutting flattening is to ensure that the pressure distribution can be uniform and better reflect the pressure distribution condition.
According to an embodiment of the present invention, as shown in fig. 2, the armrest part is implemented by using two straight plates at the left and right armrests, respectively, fixing the servo motor encoder at the fixed end 3 of the straight plate, then placing the acceleration sensor at the suspended end 1 of the straight plate, and fixing one end of the servo motor encoder at the corresponding part of the armrest of the seat by using a hinge. As shown in fig. 2: an acceleration sensor is arranged at the position of a hanging end 1 of a straight plate shown in the figure, a flexible piezoresistive sensor is laid and fixed on the upper surface of the straight plate to form a handrail component 2, a servo motor encoder is arranged at the position of a fixed end 3 at the tail end of the straight plate, the flexible piezoresistive sensor is arranged at the position of a cushion 4, and a handrail part 5 is symmetrical to a handrail part at the other side.
According to one embodiment of the present invention, it is preferable to place the multi-channel (e.g. 16 x 16) flexible piezoresistive sensors between the high density sponge and the abrasion resistant leather sandwich at the armrest portion of the chair on both sides, and to use strong adhesive to reinforce and prevent the movement. Flexible piezoresistive sensors in multiple channels (e.g., 32 x 32) are placed between a layer of abrasion resistant leather material and a high density sponge in the seat cushion area of the chair and reinforced with a high strength slow drying adhesive. The high-strength slow-drying glue can ensure no wrinkling and no deviation, thereby ensuring the quality of pressure distribution change data;
according to an embodiment of the present invention, as shown in fig. 1, the lower computer driving circuit board of the present invention is configured to acquire data of a flexible pressure sensor, and includes a 5V power module driver and two 3.3V multiplexers, and divides a flexible piezoresistive sensing single channel voltage by using a voltage dividing resistor to obtain an analog voltage signal with an appropriate size, and then performs digital-to-analog signal conversion on the analog voltage signal by using a DSP, and finally encapsulates the digital voltage signals of a plurality of nodes into frames, and transmits the frames to an upper computer through a wired serial port or transmits the frames to the upper computer through bluetooth.
The motor encoder driver adopts a servo motor driver with proper power (approximately equal to 200W) and low voltage (approximately 36V) and high current (approximately 7A), the driver and an upper computer adopt wired serial port transmission, and the transmission content comprises rotating speed and torque. The acceleration sensor adopts an acceleration sensor and is directly connected with an upper computer through a flat cable to transmit the posture data of the two handrail parts.
According to a preferred embodiment of the present invention, the flexible piezoresistive sensor is disposed between a high-density sponge and a wear-resistant leather interlayer of the chair, and is fixed by a cloth-covered double-sided adhesive tape, and a data output cable of the flexible piezoresistive sensor is connected to a circuit board fixed at a backrest of the chair, a power interface of the driving circuit board is a 5V conventional USB-type a power interface, a motor encoder is connected by solder, and a transformer module (e.g., 220V-5V, such as 220V-36V) is installed at the same time. And then the assembly of the intelligent seat is completed.
Preferably, the upper computer is realized by an ARM platform single chip microcomputer, clicking and inputting operations are performed through a touch screen, the upper computer system is a Ubuntu 16.04 system, the software system is compiled by a cross-platform QT development framework, and the programming language is C + +. The upper computer working process comprises the following steps:
confirming start/stop data transmission, receiving an original data frame of a lower computer, unpacking the original data, preprocessing the data (eliminating noise and storing data), and then visually displaying the processed data on an interface of the upper computer, as shown in fig. 4, wherein the display content comprises: the pressure of each node of the flexible piezoresistive sensor (represented by different colors) and the total pressure change line graph of the surface of the flexible piezoresistive sensor.
According to one embodiment of the invention, the parkinson patient data acquisition and processing using the aforementioned smart chair has the following improvements:
1. the data uses a computer vision algorithm which cannot be used by the traditional piezoresistive sensor, and specifically, the data is different from the original single-point sensor which is only pressure change data changing along with time, and the data of the flexible piezoresistive sensor is as follows: multiple channels (e.g. 64 × 32 — 2048 points), high refresh rate (up to 40Hz), and more information and features than single point sensing data. If the data is treated as gray video, for example, in tremor recognition, the CNN + GRU model can be used to extract and classify features.
2. Unlike conventional piezoresistive sensors, more complex pre-processing is required. For example, the collected data may be regarded as similar video data, and has a low resolution (e.g. 64 × 32 pixels), and only has the characteristic of information of pressure magnitude change, so that it may be called: grayscale low resolution type video data. Therefore, for this kind of video data, it is necessary to perform a data preprocessing process on the video data, which includes first using median filtering to eliminate possible crosstalk noise of the sensor (i.e., interference noise generated by mutual influence between conductive metal strips), using interpolation to perform upsampling to improve data (video) resolution, using a smoothing method commonly used in digital image processing to uniformly smooth the image, and filtering part of the noise.
According to one embodiment of the invention, the memory of the upper computer stores data acquisition and processing instructions, and when executed by the processor of the singlechip, the method performs the following steps to acquire and process the Parkinson's patient data:
step 1, tremor data acquisition and processing:
step 1.1: collecting data received by the flexible resistance sensors at the handrail parts at the two sides; firstly, a median filter is used for eliminating noise, an interpolation method is used for up-sampling, and then a Gaussian filter is used for smoothing to obtain preprocessed data;
step 1.2: inputting the preprocessed gray low-resolution video data into a convolutional neural network for feature dimensionality reduction extraction, inputting the extracted feature data into a GRU (generalized regression) cyclic neural network for classification, wherein a cross entropy loss function is used as a loss function:
Figure BDA0002799924990000091
Figure BDA0002799924990000092
where N is the number of samples, M is the number of classes (e.g., in this application scenario: five total levels 0-4), yicIs an indicator variable (1 if the class sample is the same as the class of sample i, otherwise 0); p is a radical oficIs the prediction probability for a prediction sample belonging to class c. After prediction, a tremor factor S is outputT；
Tremor factor STThe calculation process is as aboveWherein p iscFor the prediction probability of the model for level c, c is the weight corresponding to that level, and the weight of the lower level can be adjusted slightly higher. (for example, 0 grade in the original tremor grade corresponds to 1.2, 1 corresponds to 2.2, 2 corresponds to 2.1, 3 corresponds to 3, and 4 corresponds to 4, so that the aim of improving the conservative evaluation of medical staff on tremor invisible to the naked eye and improving the overall evaluation performance of equipment is achieved) the tremor factor has a strong relationship with the tremor performance of an evaluation object, and the tremor amplitude and frequency of the evaluation object can be better reflected.
The model training data can be from the labeled data of professionals and can also be from predefined data, and according to the severity of tremor of an evaluation object, the invention can divide the tremor into a plurality of grades and label the grades, so that the tremor can be evaluated more comprehensively, as shown in fig. 5.
According to yet another embodiment of the present invention, another method of parkinson's patient data acquisition and processing tremor assessment data is provided:
step 1.1: collecting data received by the flexible resistance sensors at the handrail parts at the two sides; firstly, filtering medium-high frequency noise in data by using a bilateral filter to obtain preprocessed data;
step 1.2: inputting each frame into a self-encoder to extract and reduce the dimension of the frame characteristics for the preprocessed gray low-resolution video data, and splicing the extracted characteristic data of each frame according to the sequence of the frame numbers to form a data matrix containing all the data characteristics; inputting the matrix into a support vector machine, judging whether a prediction error occurs or not if the data matrix is used for predicting the tremor degree, if the prediction error occurs, outputting 0 by the support vector machine, and if the prediction error does not occur, outputting 1 by the support vector machine; if the output of the data matrix obtained in the support vector machine is 1, inputting the data matrix into a convolutional neural network for time sequence feature extraction and classification, wherein the loss function uses a multi-class cross entropy function:
Figure BDA0002799924990000101
yicis an indicator variable (1 if the class sample is the same as the class of sample i, otherwise 0); p is a radical oficIs the prediction probability for a prediction sample belonging to class c. If the output of the data matrix obtained in the support vector machine is 0, the data of the batch is not classified, and the next batch of data is waited. After classifying the data, output the tremor factor:
Figure BDA0002799924990000102
wherein P isSVMFor support vector machine classification accuracy on training set, lkFor the prediction of tremor extent of this batch of data,/iDegree of tremor corresponding to category i, piM is the set of all tremor degrees, which is the predicted probability that the sample belongs to category i.
In the data processing model, the bilateral filter considers the intensity difference and the spatial relationship between data points at the same time, so that the bilateral filter can effectively filter out high-frequency and medium-frequency noise, and reserve the edge of the effective part of input data to avoid the ringing effect as much as possible. The self-encoder can extract the main features of the input data to the maximum extent, reduce the dimensionality of the input data and save computing resources for subsequent processing. The support vector machine plays the role of a data filter in this model. The method can filter the data which can be classified wrongly in the subsequent convolutional neural network model, and only the data which can be classified correctly can pass through, so that the accuracy of the model prediction can be improved. Convolutional layers in convolutional neural networks are used to extract timing features in the data, and fully-connected layers classify the data according to the timing features.
Step 2, muscle stiffness assessment data acquisition and processing
Step 2.1: collecting data received by the flexible resistance sensors at the handrail parts at the two sides; firstly, carrying out the same data preprocessing in the step 1.1;
step 2.2: summing the pressures obtained at the flexible resistance sensors at the positions of the single-side handrails, namely obtaining the total pressure at the positions of the handrails through the data of 16 channels by 16 channels; linear regression is performed by the encoder angle and the modulus of the acceleration sensor in three linear directions:
Figure BDA0002799924990000111
Figure BDA0002799924990000112
wherein, R is the rigidity grade (0-4 grade) in the UPDRS scale;
when outputting, R0 may be removed here, as long as the latter part serves as a strong factor as a characteristic value, that is, outputting:
Figure BDA0002799924990000113
the output value is a strong direct factor (multi-sensor fusion value), R0Alpha, beta and lambda are regression variables, P is the total pressure value of the handrail, S is the rotation angle of the encoder,
Figure BDA0002799924990000114
is the mode of the three-dimensional accelerometer.
Step 3, acquiring and processing data in the rising stage
The rising data acquisition stage is divided into three stages: a gravity center transfer stage, a foot supporting stage and a separating rising stage; and (4) calculating the total pressure value and the rising time of each stage of the three stages respectively.
Three-stage division:
A. a gravity center transfer stage: the human body is supported by the armrests, the gravity center is gradually transferred from the buttocks to the feet, but the buttocks still have gravity. The behavior on the flexible pressure sensor data is as follows: the total pressure value of the sensor of the cushion part is reduced, the gravity center is changed, and the stage can be judged by setting a gravity center change threshold value.
B. A step supporting stage: the human body is supported by the armrests, the buttocks are separated from the chair, and no gravity exists in the cushion part. The expression on the flexible pressure sensor data is that the total pressure value of the sensor of the cushion part is zero, the total pressure sensor of the armrest part is obviously increased compared with the first stage, and the first stage and the second stage can be distinguished by judging whether the gravity of the cushion part exists or not.
C. A separation rising stage: the human body is supported by the feet, and no gravity exists in the seat cushion and the armrest part in the stage of gradually standing. The second and third stages can be distinguished by determining whether pressure is present on the armrest.
Specifically, the feature calculation at each stage is as follows:
1) total pressure value:
Figure BDA0002799924990000121
calculating the total pressure of a frame in the video-like data, wherein: i, j are pixel points (sensing points) positioned in i rows and j columns, m, n are the row number and column number of the matrix, AijI rows and j columns of pressure values.
2) Change of center of gravity:
firstly, a Cartesian coordinate system is established for a frame, for the condition that the pressure in the frame is greater than a specific value, the gravity x coordinates of all pixel points (sensing points) are added and divided by the number of the pixel points, so that the gravity center position in the x direction can be obtained, and similarly, the gravity center position in the y direction can also be obtained, and if the change value of the gravity center position in the set direction exceeds a set threshold value, the first stage can be judged.
3) Duration of the phase: the user (medical staff) clicks the start of the getting-up evaluation as the initial frame of the first stage, and records the corresponding time stamp, when the first stage gravity center change threshold frame is reached, the frame is marked as the end frame of the first stage, the corresponding time stamp is recorded, and the stage duration is obtained by utilizing the two time stamps. The last two stages are as above.
Step 4, collecting and processing data on two sides of the body:
the tremor factor data, the muscle strength data and the maximum pressure conditions of the armrests on the two sides in the rising process of the patient are weighted and calculated, and the unbalance degree of the limbs on the two sides can be obtained:
S＝aST+bSR+cP
wherein S is a single-side limb index, a, b and c are weights of each parameter, STIs tremor factor, SRMuscle rigidity factor, P is the maximum pressure of the armrest in the rising stage; finally, the left index is used for dividing the right index to obtain the difference degree (%) of the bilateral limbs;
Figure BDA0002799924990000122
although illustrative embodiments of the present invention have been described above to facilitate the understanding of the present invention by those skilled in the art, it should be understood that the present invention is not limited to the scope of the embodiments, but various changes may be apparent to those skilled in the art, and it is intended that all inventive concepts utilizing the inventive concepts set forth herein be protected without departing from the spirit and scope of the present invention as defined and limited by the appended claims.

Claims (11)

1.一种多功能数字化评估智能座椅，其特征在于，包括：座椅本体、底座、坐垫部位、扶手部位；1. A multifunctional digital evaluation intelligent seat, characterized in that, comprising: a seat body, a base, a seat cushion part, and an armrest part; 在智能座椅坐垫部位、以及扶手部位分别安装柔性压阻传感器，在扶手部位加装伺服电机编码器以及加速度传感器；Install flexible piezoresistive sensors on the seat cushion and armrests of the smart seat respectively, and install servo motor encoders and acceleration sensors on the armrests; 所述柔性压阻传感器用于获取表面上方接触面压力分布变化数据；所述柔性压阻传感器包括：两层导电层，所述导电层中设置有不锈钢导电纤维与聚乙烯纤维；其中，两层导电层呈九十度水平垂直排列，并在两层导电层中间放置一层柔性半导体压敏电阻层，所述压敏电阻层为碳化硅纤维与聚乙烯纤维交错排列组成；随后用缝纫机将三层车线固定，并使用热熔胶进行加固；The flexible piezoresistive sensor is used to obtain the change data of the pressure distribution of the contact surface above the surface; the flexible piezoresistive sensor includes: two layers of conductive layers, and the conductive layers are provided with stainless steel conductive fibers and polyethylene fibers; wherein, the two layers The conductive layers are arranged horizontally and vertically at 90 degrees, and a flexible semiconductor varistor layer is placed between the two conductive layers. The varistor layer is composed of silicon carbide fibers and polyethylene fibers staggered; The layer car line is fixed and reinforced with hot melt adhesive; 所述伺服电机编码用于获取手臂旋转数据，扶手部位加速度传感器用于测量手臂的震动数据；The servo motor code is used to obtain the arm rotation data, and the acceleration sensor of the armrest is used to measure the vibration data of the arm; 上位机，包括存储器及单片机处理器；The host computer, including the memory and the single-chip processor; 上述各数据通过有线串口协议传输，由下位机发送至上位机，并在上位机上利用单片机处理器对数据进行处理分析，得到被采集病患的相关参数，并将数据与分析结果进行存储和显示；The above data is transmitted through the wired serial port protocol, and sent from the lower computer to the upper computer, and the data is processed and analyzed by the single-chip processor on the upper computer to obtain the relevant parameters of the collected patients, and the data and analysis results are stored and displayed. ; 所述的上位机中的存储器中存储有数据采集和处理指令，被单片机处理器执行时，执行以下步骤以进行帕金森患者数据采集和处理：The memory in the host computer stores data acquisition and processing instructions, and when executed by the single-chip processor, executes the following steps to collect and process Parkinson's patient data: 步骤1、震颤数据采集与处理；Step 1. Tremor data collection and processing; 步骤2、肌肉僵直数据采集处理；Step 2. Muscle stiffness data collection and processing; 步骤3、起身阶段数据采集与处理；Step 3. Data collection and processing in the getting-up stage; 步骤4、身体双侧数据采集和处理；Step 4. Data collection and processing on both sides of the body; 所述步骤3起身阶段数据采集与处理具体包括：The data collection and processing of the step 3 getting up stage specifically includes: 起身数据采集阶段划分为三个阶段：重心转移阶段、脚部支撑阶段、分离起身阶段；对于三个阶段，分别对其各个阶段总压力值、起身时间进行计算；The data collection stage of getting up is divided into three stages: the center of gravity transfer stage, the stage of foot support, and the stage of separation and getting up; for the three stages, the total pressure value and the time to get up in each stage are calculated respectively; 三个阶段的划分：The division of three stages: A.重心转移阶段：人体借助扶手支撑，将重心由臀部逐渐转移至脚部，但臀部仍有重力存在，在柔性压感传感器数据上的表现为：坐垫部分传感器总压力值减少，重心发生变化，可通过设定重心变化阈值判断此阶段；A. Center of gravity transfer stage: With the support of the armrest, the human body gradually transfers the center of gravity from the hips to the feet, but the hips still have gravity. The data of the flexible pressure sensor is as follows: the total pressure value of the sensor on the seat cushion decreases, and the center of gravity changes. , this stage can be judged by setting the change threshold of the center of gravity; B.脚步支撑阶段：人体借助扶手支撑，臀部脱离椅子，坐垫部分无重力存在；在柔性压感传感器数据上的表现为，坐垫部分传感器总压力值为零，扶手部分总压力传感器相较于第一阶段有明显增加，可通过判断坐垫部分重力是否存在区分第一阶段与第二阶段；B. Step support stage: the human body is supported by the armrests, the buttocks are separated from the chair, and the seat cushion part has no gravity; the performance of the flexible pressure sensor data is that the total pressure value of the cushion part sensor is zero, and the total pressure sensor of the armrest part is compared with the first part. There is a significant increase in the first stage, and the first stage and the second stage can be distinguished by judging whether the gravity of the seat cushion part exists; C.分离起身阶段：人体依靠脚部支撑，逐渐站立的阶段，坐垫、扶手部分均无重力存在，可通过判断扶手是否存在压力从而区分第二、第三阶段；C. Separation and getting up stage: The human body relies on the support of the feet and gradually stands up. The seat cushion and armrest have no gravity. The second and third stages can be distinguished by judging whether there is pressure on the armrest; 具体的，各个阶段特征计算如下：Specifically, the characteristics of each stage are calculated as follows: 1)总压力值：1) Total pressure value:
Figure FDA0003166447260000021
Figure FDA0003166447260000021
 计算类视频数据中一帧的总压力，其中：i，j为位于i行j列的像素点，即传感点，m，n为矩阵的行数、列数，Aij为i行j列压力值；Calculate the total pressure of a frame in video-like data, where: i, j are the pixels located in row i, column j, that is, the sensing point, m, n are the number of rows and columns of the matrix, and A ij is row i and column j Pressure value; 2)重心变化：2) The center of gravity changes: 首先对一帧建立笛卡尔坐标系，对于一帧中压力大于特定值的情况，将所有像素点的重力x坐标相加并除以像素点个数，即得到x方向重心位置，同理，得y方向重心位置，若设定方向重心位置变化值超过设定阈值，即可判定为第一阶段；First, a Cartesian coordinate system is established for a frame. For the case where the pressure in a frame is greater than a certain value, the gravity x coordinates of all pixels are added and divided by the number of pixels, that is, the position of the center of gravity in the x direction is obtained. The position of the center of gravity in the y direction, if the change in the position of the center of gravity in the set direction exceeds the set threshold, it can be determined as the first stage; 阶段持续时间：用户点击起身评测开始，作为第一阶段的起始帧，并记录相应时间戳，当达到第一阶段重心变化阈值帧时，标记为第一阶段的结束帧，记录相应时间戳，利用两时间戳得到阶段持续时间。Stage duration: The user clicks to get up and the evaluation starts, as the start frame of the first stage, and the corresponding timestamp is recorded. When the threshold frame of the center of gravity change of the first stage is reached, it is marked as the end frame of the first stage, and the corresponding timestamp is recorded. Use the two timestamps to get the phase duration. 2.根据权利要求1所述的一种多功能数字化评估智能座椅，其特征在于，通过将不锈钢纤维纺织为金属纱线，并将其与普通纱线即聚乙烯纱线交错纺织为多条金属纱线条纹构成导电层，将两层导电层垂直叠放构成矩阵网格排列，并在中间放置柔性半导体压敏电阻层，并将矩阵网格各条纹金属带通过良导体进行连接，并将其与外部导线连接导通，制作得到柔性压阻传感器。2. A multifunctional digital evaluation smart seat according to claim 1, characterized in that, by weaving stainless steel fibers into metal yarns, and interlacing them with ordinary yarns, that is, polyethylene yarns, they are woven into a plurality of strips. The metal yarn stripes form the conductive layer, the two conductive layers are vertically stacked to form a matrix grid arrangement, and a flexible semiconductor varistor layer is placed in the middle, and the striped metal strips of the matrix grid are connected by good conductors, and the The flexible piezoresistive sensor is manufactured by connecting it with the external wire and conducting. 3.根据权利要求1所述的一种多功能数字化评估智能座椅，其特征在于，所述智能座椅包括由不锈钢圆盘底座、增压型油泵液压升降杆组成的底座部分，椅体部分包括由强化复合直板与高密度海绵组成的构件，外表覆盖耐磨皮革材料所述不锈钢圆盘底座为圆形旋转液压底座；将圆形旋转液压底座与座椅垫片通过点焊连接，并将强化复合直板与切削平整的高密度海绵粘连形成各个构件，包括扶手、靠背、坐垫构件，使用钢钉将各构件固定连接，最后表面覆盖耐磨皮革材料进行封装。3. A multifunctional digital evaluation smart seat according to claim 1, wherein the smart seat comprises a base part consisting of a stainless steel disc base and a hydraulic lifting rod of a pressurized oil pump; It includes a component composed of reinforced composite straight plate and high-density sponge, and the surface is covered with wear-resistant leather material. The stainless steel disc base is a circular rotary hydraulic base; the circular rotary hydraulic base and the seat gasket are connected by spot welding, and the The reinforced composite straight plate and the flat-cut high-density sponge are adhered to form various components, including armrests, backrests, and seat cushion components. Steel nails are used to fix and connect each component, and finally the surface is covered with wear-resistant leather material for packaging. 4.根据权利要求1所述的一种多功能数字化评估智能座椅，其特征在于，在左右扶手处，分别采用两块直板，将伺服电机编码器固定在直板固定端，然后将加速度传感器放置在直板悬空端，并采用铰链将伺服电机编码器一端固定在座椅扶手相应部位，将柔性压阻传感铺设固定在直板上表面，形成扶手构件，坐垫位置安装柔性压阻传感器，所述左右扶手构件对称。4. A multifunctional digital evaluation smart seat according to claim 1, characterized in that, at the left and right armrests, two straight plates are respectively used, the servo motor encoder is fixed on the fixed end of the straight plate, and then the acceleration sensor is placed At the suspended end of the straight plate, one end of the servo motor encoder is fixed to the corresponding part of the armrest of the seat by a hinge, and the flexible piezoresistive sensor is laid and fixed on the surface of the straight plate to form an armrest member, and the flexible piezoresistive sensor is installed at the seat cushion. The armrest members are symmetrical. 5.根据权利要求1所述的一种多功能数字化评估智能座椅，其特征在于，将多通道的柔性压阻传感器安置于两侧椅子扶手部位的高密度海绵与耐磨皮革材料夹层中间，并使用粘合剂进行加固，防止其移动；将多通道的柔性压阻传感器安置于椅子坐垫部位的耐磨皮革材料与海绵夹层中间，并使用慢干型粘合剂进行加固，保证不起皱、不偏移，从而保证压力分布变化数据质量。5. A multi-functional digital evaluation intelligent seat according to claim 1, wherein the multi-channel flexible piezoresistive sensor is placed between the high-density sponge and the wear-resistant leather material sandwiched between the armrests of the chair on both sides, It is reinforced with adhesive to prevent it from moving; the multi-channel flexible piezoresistive sensor is placed between the wear-resistant leather material and the sponge interlayer on the seat cushion of the chair, and reinforced with slow-drying adhesive to ensure no wrinkle , not offset, so as to ensure the data quality of pressure distribution changes. 6.根据权利要求1所述的一种多功能数字化评估智能座椅，其特征在于，下位机驱动电路板，用于采集柔性压阻传感器数据，所示下位机驱动电路板包括电源模块驱动、两个多路选择器，并使用分压电阻对柔性压阻传感单道电压进行分压，得到模拟电压信号，后使用DSP对模拟电压信号进行数字模拟信号转换，最终将多个结点的数字电压信号封装成帧，通过有线串口传输至上位机。6 . The multifunctional digital evaluation intelligent seat according to claim 1 , wherein the lower computer drive circuit board is used to collect data of flexible piezoresistive sensors, and the shown lower computer drive circuit board comprises a power module drive, Two multiplexers, and use a voltage divider to divide the voltage of the flexible piezoresistive sensing single-channel to obtain an analog voltage signal, and then use the DSP to convert the analog voltage signal to digital to analog signal, and finally convert the voltage of multiple nodes. The digital voltage signal is encapsulated into a frame and transmitted to the host computer through the wired serial port. 7.根据权利要求1所述的一种多功能数字化评估智能座椅，其特征在于，电机编码器驱动器采用安全电压下的伺服电机驱动器，驱动器与上位机采用有线传输，传输内容包括转速与扭矩，加速度传感器采用加速度传感器，直接通过排线与上位机连接传输两扶手部位姿态数据。7. A multifunctional digital evaluation intelligent seat according to claim 1, wherein the motor encoder driver adopts a servo motor driver under safe voltage, and the driver and the host computer adopt wired transmission, and the transmission content includes rotational speed and torque , The acceleration sensor adopts the acceleration sensor, which is directly connected to the host computer through the cable to transmit the attitude data of the two armrests. 8.根据权利要求7所述的一种多功能数字化评估智能座椅，其特征在于，所述步骤1震颤数据采集与处理具体包括：8. The multifunctional digital evaluation intelligent seat according to claim 7, wherein the step 1 tremor data collection and processing specifically includes: 步骤1.1：采集两侧扶手部位柔性电阻传感器接收到的数据；首先使用中值滤波器进行噪声消除，并采用插值法进行上采样，随后使用高斯滤波器进行平滑处理，得到预处理后数据；Step 1.1: Collect the data received by the flexible resistance sensors on the handrails on both sides; first use the median filter to remove noise, and use the interpolation method for upsampling, and then use the Gaussian filter for smoothing to obtain the preprocessed data; 步骤1.2：对于预处理后灰度低分辨率类视频数据，输入卷积神经网络进行特征降维提取，将提取后的特征数据输入GRU循环神经网络中进行分类，损失函数使用交叉熵损失函数：Step 1.2: For the preprocessed grayscale low-resolution video data, input the convolutional neural network for feature dimension reduction extraction, and input the extracted feature data into the GRU recurrent neural network for classification. The loss function uses the cross entropy loss function:
Figure FDA0003166447260000031
Figure FDA0003166447260000031
Figure FDA0003166447260000032
Figure FDA0003166447260000032
 其中，N为样本的数量，M为类别的数量，yic为指示变量；pic为对于预测样本属于类别c的预测概率，预测后，输出震颤因子ST，震颤因子ST计算过程如上，其中pc为模型对于级别c的预测概率，c为对应该级别的权值。Among them, N is the number of samples, M is the number of categories, and y ic is an indicator variable; pic is the predicted probability that the predicted sample belongs to category c. After the prediction, the tremor factor S T is output, and the calculation process of the tremor factor S T is as above, where p c is the predicted probability of the model for level c, and c is the weight corresponding to the level. 9.根据权利要求1所述的一种多功能数字化评估智能座椅，其特征在于，步骤1震颤数据采集与处理包括：9. The multi-functional digital evaluation intelligent seat according to claim 1, wherein the tremor data collection and processing in step 1 comprises: 步骤1.1：采集两侧扶手部位柔性电阻传感器接收到的数据；首先使用双边滤波器过滤掉数据中的中高频噪声，得到预处理后数据；Step 1.1: Collect the data received by the flexible resistance sensors at the handrails on both sides; first, use a bilateral filter to filter out the medium and high frequency noise in the data to obtain the preprocessed data; 步骤1.2：对于预处理后灰度低分辨率类视频数据，将每一帧输入到自编码器中进行该帧特征的提取与降维，然后将每一帧提取后的特征数据按照帧数顺序进行拼接，组成一个包含所有数据特征的数据矩阵；将该矩阵输入到支持向量机中，判断若将该数据矩阵用于预测震颤程度，是否会预测错误，若会预测错误，支持向量机输出为0，若不会预测错误，支持向量机输出为1；若该数据矩阵在支持向量机中获得的输出为1，则将该数据矩阵输入到卷积神经网络中进行时序特征提取以及分类，损失函数使用多类别交叉熵函数：Step 1.2: For the preprocessed grayscale low-resolution video data, input each frame into the auto-encoder to extract and reduce the feature of the frame, and then extract the feature data of each frame in the order of the number of frames Splicing to form a data matrix containing all data features; input the matrix into the support vector machine, and judge whether the data matrix is used to predict the degree of tremor, whether the prediction will be wrong, if the prediction is wrong, the output of the support vector machine is 0, if there is no prediction error, the output of the support vector machine is 1; if the output of the data matrix obtained in the support vector machine is 1, the data matrix is input into the convolutional neural network for time series feature extraction and classification, and the loss is The function uses the multi-class cross-entropy function:
Figure FDA0003166447260000041
Figure FDA0003166447260000041
 yic为指示变量，若此类别样本与样本i的类别相同则为1，否则为0；pic为对于预测样本属于类别c的预测概率；若该数据矩阵在支持向量机中获得的输出为0，则不对这批数据进行分类，并等待下一批数据；对数据进行分类后，输出震颤因子：y ic is the indicator variable, if the category of the sample is the same as that of the sample i, it is 1, otherwise it is 0; pic is the predicted probability that the predicted sample belongs to category c; if the data matrix is obtained in the support vector machine The output is 0, do not classify this batch of data and wait for the next batch of data; after classifying the data, output the tremor factor:
Figure FDA0003166447260000042
Figure FDA0003166447260000042
 其中PSVM为支持向量机在训练集上的分类正确率，lk为该批数据的预测震颤程度，li为类别i对应的震颤程度，pi为样本属于类别i的预测概率，M为所有震颤程度的集合。where P SVM is the classification accuracy rate of the support vector machine on the training set, l k is the predicted tremor degree of the batch of data, li is the tremor degree corresponding to category i , pi is the predicted probability that the sample belongs to category i , and M is the A collection of all tremor levels. 10.根据权利要求9所述的一种多功能数字化评估智能座椅，其特征在于，所述步骤2肌肉僵直数据采集处理具体包括：10. The multifunctional digital assessment intelligent seat according to claim 9, wherein the step 2 data collection and processing of muscle stiffness specifically includes: 步骤2.1采集两侧扶手部位柔性电阻传感器接收到的数据；首先进行步骤1.1相同的数据预处理；Step 2.1 Collect the data received by the flexible resistance sensors on the handrails on both sides; first perform the same data preprocessing as in step 1.1; 步骤2.2：对于单侧扶手部位柔性电阻传感器处得到压力进行求和，即16*16通道的数据，得到扶手处总压力；由编码器角度、加速度传感器三个线性方向的加速度的模进行线性回归：Step 2.2: Sum the pressure obtained at the flexible resistance sensor of the one-sided armrest, that is, the data of 16*16 channels, to obtain the total pressure at the armrest; linear regression is performed by the modulo of the acceleration in the three linear directions of the encoder angle and the acceleration sensor :
Figure FDA0003166447260000043
Figure FDA0003166447260000043
Figure FDA0003166447260000051
Figure FDA0003166447260000051
 其中，R为UPDRS量表中强直等级；Among them, R is the tonic grade in the UPDRS scale; 在输出时，将R0去掉，只要后面的部分作为强直因子作为特征值，即输出：When outputting, remove R 0 , as long as the latter part is used as a tonic factor as an eigenvalue, that is, output:
Figure FDA0003166447260000052
Figure FDA0003166447260000052
 输出值为强直因子为多传感器融合值，R0、α、β、λ为回归变量，P为扶手总压力值，S为编码器旋转角度，
Figure FDA0003166447260000053
为三维加速度计的模。The output value of the tonic factor is the multi-sensor fusion value, R 0 , α, β, λ are the regression variables, P is the total pressure value of the armrest, S is the encoder rotation angle,
Figure FDA0003166447260000053
is the modulus of the three-dimensional accelerometer. 11.根据权利要求1所述的一种多功能数字化评估智能座椅，其特征在于，所述步骤4身体双侧数据采集和处理具体包括：11. The multi-functional digital assessment smart seat according to claim 1, wherein the step 4 of collecting and processing data on both sides of the body specifically includes: 对于患者双侧扶手得到的震颤因子数据、肌肉强直数据、以及起身过程中两侧扶手的最大压力情况进行加权计算，可得到两侧肢体不平衡程度：The tremor factor data, muscle rigidity data, and the maximum pressure of the armrests on both sides of the patient during the process of getting up are weighted to calculate the degree of imbalance of the limbs on both sides: S＝aST+bSR+cPS=aS T +bS R +cP 其中，S为单侧肢体指标，a、b、c为各参数权重，ST为震颤因子、SR为肌肉强直因子、P为起身阶段扶手最大压力；最终，使用左侧指标除右侧指标，得到双侧肢体差异程度(％)Among them, S is the unilateral limb index, a, b, and c are the weights of each parameter, S T is the tremor factor, S R is the muscle rigidity factor, and P is the maximum pressure of the armrest in the getting-up stage; finally, the left index is used to divide the right index , to get the degree of difference in bilateral limbs (%)
Figure FDA0003166447260000054
Figure FDA0003166447260000054
CN202011345922.8A 2020-11-26 2020-11-26 A multifunctional digital evaluation intelligent seat Active CN112315459B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011345922.8A CN112315459B (en) 2020-11-26 2020-11-26 A multifunctional digital evaluation intelligent seat

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011345922.8A CN112315459B (en) 2020-11-26 2020-11-26 A multifunctional digital evaluation intelligent seat

Publications (2)

Publication Number Publication Date
CN112315459A CN112315459A (en) 2021-02-05
CN112315459B true CN112315459B (en) 2021-10-22

Family

ID=74307926

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011345922.8A Active CN112315459B (en) 2020-11-26 2020-11-26 A multifunctional digital evaluation intelligent seat

Country Status (1)

Country Link
CN (1) CN112315459B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113358249B (en) * 2021-06-11 2022-09-30 中国科学技术大学 Fabric type piezoresistive sensor array and intelligent object
US20230054076A1 (en) * 2021-08-18 2023-02-23 Advanced Neuromodulation Systems, Inc. Systems and methods for providing digital health services
CN114323368A (en) * 2021-12-09 2022-04-12 深圳先进技术研究院 Flexible intelligent sitting posture monitoring system based on hip pressure

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105426696A (en) * 2015-12-24 2016-03-23 中国科学院苏州生物医学工程技术研究所 Multi-node quantitative assessment system and method for symptoms of Parkinson's disease
CN108814062A (en) * 2018-06-20 2018-11-16 上海掌门科技有限公司 Multifunctional seat and multifunctional seat system

Family Cites Families (20)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040260156A1 (en) * 2001-05-18 2004-12-23 Commwell, Inc. Chair and ancillary apparatus with medical diagnostic features in a remote health monitoring system
US6852086B2 (en) * 2001-06-18 2005-02-08 Dan Atlas Detection of signs of attempted deception and other emotional stresses by detecting changes in weight distribution of a standing or sitting person
US10258259B1 (en) * 2008-08-29 2019-04-16 Gary Zets Multimodal sensory feedback system and method for treatment and assessment of disequilibrium, balance and motion disorders
WO2011133799A1 (en) * 2010-04-21 2011-10-27 Northwestern University Medical evaluation system and method using sensors in mobile devices
EP3240459B1 (en) * 2014-12-29 2021-07-14 Herman Miller, Inc. System architecture for office productivity structure communications
WO2017048977A1 (en) * 2015-09-15 2017-03-23 The Arizona Board Of Regents On Behalf Of The University Of Arizona Smart carpet systems and methods of using same for monitoring physical and physiological activities
CN105561567B (en) * 2015-12-29 2018-11-13 中国科学技术大学 A kind of meter step and motion state apparatus for evaluating
CN105955487B (en) * 2016-05-24 2018-10-02 天津锡跃科技有限公司 VR coordinative composition of equipments seats
CN106092389A (en) * 2016-05-27 2016-11-09 电子科技大学 A kind of novel array-type flexible pressure transducer
WO2017213066A1 (en) * 2016-06-08 2017-12-14 日本電気株式会社 Tremor detection device, stress evaluation system using same, and stress evaluation method
CN107232822A (en) * 2016-08-31 2017-10-10 浙江玛拉蒂智能家具科技有限公司 A kind of Intelligent seat based on gesture recognition, intelligent interactive system and method
US10390752B2 (en) * 2016-09-23 2019-08-27 International Business Machines Corporation Cognitive chair for monitoring Parkinson's disease progression
EP3461403A1 (en) * 2017-09-29 2019-04-03 Koninklijke Philips N.V. A method and apparatus for assessing the mobility of a subject
CN108814616B (en) * 2018-04-12 2021-11-05 深圳和而泰数据资源与云技术有限公司 Sitting posture identification method and intelligent seat
CN108402755A (en) * 2018-05-16 2018-08-17 杨有冬 It is a kind of that there is the executive chair for stablizing load carrier
CN209360118U (en) * 2018-10-19 2019-09-10 国家体育总局体育科学研究所 A kind of adjustable Intelligent force measuring chair of backrest
US20200129109A1 (en) * 2018-10-30 2020-04-30 Jingbo Zhao Mobility Assessment Tracking Tool (MATT)
CN109770597A (en) * 2019-03-12 2019-05-21 中国科学技术大学 a wearable seat
CN110495741A (en) * 2019-07-25 2019-11-26 安徽永顺信息科技有限公司 A kind of Intelligent office seat system and its reminding method
CN210930422U (en) * 2019-12-09 2020-07-07 湖北工业大学 Chair with adjustable height

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105426696A (en) * 2015-12-24 2016-03-23 中国科学院苏州生物医学工程技术研究所 Multi-node quantitative assessment system and method for symptoms of Parkinson's disease
CN108814062A (en) * 2018-06-20 2018-11-16 上海掌门科技有限公司 Multifunctional seat and multifunctional seat system

Also Published As

Publication number Publication date
CN112315459A (en) 2021-02-05

Similar Documents

Publication Publication Date Title
CN112315459B (en) A multifunctional digital evaluation intelligent seat
CN104665849B (en) A kind of high ferro dispatcher based on the interaction of many physiological signals multi-model stress detection method
US11147500B2 (en) Electrocardiogram processing system for delineation and classification
Tabassian et al. Diagnosis of heart failure with preserved ejection fraction: machine learning of spatiotemporal variations in left ventricular deformation
EP3692903A1 (en) Artificial intelligence self-learning-based automatic electrocardiography analysis method and apparatus
CN110946566B (en) Heart beat classification method, device, equipment and storage medium based on U-Net network
US20200138306A1 (en) Feature selection for cardiac arrhythmia classification and screening
CN202027572U (en) Dynamic recording and analyzing device for electrocardiogram and movement
JP2023544242A (en) Electrocardiogram processing system for detecting and/or predicting cardiac events
CN112353393B (en) An intelligent driving vehicle occupant state detection system
CN111261277A (en) Heart disease detection method based on deep learning model
WO2021164346A1 (en) Method and device for predicting blood pressure
Edenbrandt et al. Neural networks for classification of ECG ST-T segments
CN105701806B (en) Depth image-based Parkinson tremor motion characteristic detection method and system
CN112932499A (en) Network training and single-lead-connection electrocardiogram data processing method, computer device and medium
CN101052967A (en) Method of medical monitoring
CN112819818B (en) Image recognition module training method and device
CN101163444B (en) Patient monitoring system
CN115177262A (en) Heart sound and electrocardiogram combined diagnosis device and system based on deep learning
CN112971800B (en) Deep learning-based electrocardio premature beat signal processing method, module and device
JPH1156818A (en) Input device
Hewson et al. Classification of elderly as fallers and non-fallers using centre of pressure velocity
CN106974644A (en) A kind of electrocardiogram R wave detection method and system
EP3265937B1 (en) Systems for displaying medical data
US20240277301A1 (en) System and a method for electrocardiographic prediction of computed tomography-based high coronary calcium score (cac)

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant