CN108965625B - An automatic pager based on brain-computer interface and its training method - Google Patents

Abstract

The invention relates to an automatic pager based on a brain-computer interface and a training method thereof, belonging to the field of human-computer interaction. The invention comprises a wireless earphone and a pager. When the user has the call demand under a certain condition, can wear wireless earphone, open the earphone switch can, wireless earphone will be in user's brain electricity collection under this condition and send to the calling set through the bluetooth, and the calling result that the calling set produced is analyzed out according to the eigenvalue of brain electricity signal to its inside brain computer interface module. The invention combines the brain-computer interface and the neural network, so that the automation degree of the pager is increased, and meanwhile, great convenience is provided for users, especially users who are not convenient to move. The device has the advantages of convenience and rapidness, and is quite simple to use.

Description

An automatic pager based on brain-computer interface and its training method

technical field

The invention relates to an automatic pager based on a brain-computer interface and a training method thereof, belonging to the technical field of human-computer interaction.

Background technique

As a new communication device, brain-computer interface has attracted much attention from all walks of life in recent years. Brain-computer interface devices can better transmit information from the brain directly to the outside world, and this process can be completed without the involvement of limbs and muscles. It can analyze the EEG signal of the brain and directly act on the external device, and the external device can complete a certain task according to the instructions of the EEG signal. Especially in the field of rehabilitation, its application and research are more extensive.

RBF neural network has a wide range of applications in many aspects, such as prediction and classification. Its mapping ability is relatively strong, and when its number of layers is appropriate and the amount of training data is sufficient, it can approximate a certain nonlinear function.

Most of the traditional pagers require a hand-held pager or a caller to make a call to a call station, but it is somewhat inconvenient for people who are inconvenient to move. In addition, when encountering a dangerous situation, the user can automatically call for help through the wireless headset in such a dangerous situation. The traditional pager shows great inconvenience and is difficult to meet the needs of life.

SUMMARY OF THE INVENTION

The invention provides an automatic pager based on a brain-computer interface, which is used to solve the problem that the traditional pager is inconvenient to use for people with disabilities.

The technical solution of the present invention is as follows: a pager based on a brain-computer interface, including a wireless earphone and a pager, the pager includes a microprocessor with a display screen and buttons on its surface; the microprocessor is used to transmit the wireless earphone. The received EEG signal is preprocessed, feature extraction, feature classification, and the decision information of the brain-computer interface is transmitted to the called object through the 4G network module; the display screen and buttons jointly complete the input and display work. The wireless earphone is used to collect the EEG signals generated by the user in a certain environment.

Further, the pager includes a microprocessor with a display screen and buttons on its surface; the microprocessor is respectively connected with a preprocessing module, a feature extraction module, a feature classification module, a Bluetooth receiving device, a power supply module, a 4G network module, storage modules are connected. The preprocessing module is used to perform denoising processing and signal amplification on the EEG signal transmitted by the wireless earphone; the feature extraction module is used for the maximum peak value, the minimum peak value, the EEG signal fluctuation time of the received EEG signal, etc. The characteristic value of the device is analyzed and extracted; the feature classification module is used to use the rbf neural network algorithm to determine the event characteristic value of the user call; the power module is used to provide power for the normal operation of the pager; the bluetooth receiving device is used to receive The EEG signal sent by the wireless headset; the storage module is used to store the information set by the user; the 4G network module is used to communicate with the object called by the user. The user character feature value setting module is used to set the user's introverted and extroverted characteristics; the call object selection module is used for the user to select the object to be called; the called object information storage module is used to store some information of the called object.

Further, the wireless headset includes a switch module and an electrode module integrating a Bluetooth switch, a power switch, and an EEG acquisition switch; the switch module integrating a Bluetooth switch, a power switch, and an EEG acquisition switch is used for Bluetooth selection, The functions of the power switch and the EEG acquisition switch; the electrode module is used to sense the EEG signals generated by the brain on the head.

The training method of the above-mentioned automatic pager based on the brain-computer interface is as follows: First, the tester is required to make calls in ordinary daily life, and the EEG signal of the tester is collected in the ordinary daily life by using a wireless headset, and the feature extraction module is used to collect the EEG signals of the tester. Extract the characteristic value of the tester's EEG signal at this time, that is, the maximum peak value, the minimum peak value, and the time value of the large fluctuation of the EEG signal; and then set the character characteristic value of the tester on the pager; Group training data, the input data is the maximum peak value, the minimum peak value, the time value of EEG signal fluctuations, the tester's character characteristic value, the output data is the characteristic value corresponding to the general situation, which is 1; in general, test at least 500 different testers. Secondly, perform the test in an emergency situation, extract the characteristic value of the tester's EEG signal in the emergency situation, and set the tester's personality characteristic value, the output in an emergency situation is 2, and test at least 500 different tests By. Finally, perform the test in an ultra-emergency situation, extract the tester's EEG characteristic value in the ultra-emergency situation, and set the tester's character characteristic value, the output in the ultra-emergency situation is 3, and test at least 500 different testers. According to at least 1500 sets of input and output data obtained from the above tests, the rbf neural network of the feature classification module of the pager is trained, and the structure of the rbf neural network is shown in Figure 2. The rbf neural network has a hidden layer, an input layer, and an output layer. The output of the hidden layer is shown in formula (1).

In formula (1) i∈[1 10], represents 10 neurons in the first hidden layer; X _c represents the center value of the Gaussian function, which is a sequence vector of the same dimension as X; σ _i is the Gaussian The width of the function; X is the input vector group, as shown in formula (2):

The output of the network is thus obtained as shown in formula (3):

The rbf is trained according to at least 1500 sets of data provided above, and the final σ _i , X _c , and ω _i1 are obtained by at least 1500 iterations of training according to the gradient descent method.

Beneficial effects of the invention: the pager can set the character characteristic value and the object the user wants to call according to the user's character characteristic, and by combining with the brain-computer interface, great convenience is provided for the user.

Description of drawings

Fig. 1 is the schematic diagram of pager;

Fig. 2 is the structure diagram of rbf neural network for feature classification module;

Fig. 3 is the working principle diagram of the pager after the training is completed.

Detailed ways

The content of the invention will be described in detail below with reference to specific embodiments and accompanying drawings.

Embodiment 1: The automatic pager includes a wireless headset and a pager. As shown in Figure 1, it includes a wireless headset and a pager, and the pager has a microprocessor inside, and the microprocessor is respectively connected with a Bluetooth receiving device, a preprocessing module, a feature extraction module, a feature classification module, a 4G network module, The power supply module is connected with the storage module, and the surface is provided with a display screen and button buttons. The corresponding operation can be completed according to the corresponding prompts on the display screen. The button buttons are connected with the internal microprocessor; The preprocessing module is used for denoising and signal amplification of the received EEG signal; the feature extraction module is used for the maximum peak value, the minimum peak value and the EEG signal of the received EEG signal The large fluctuation time is analyzed and extracted; the power module is used to provide electrical energy for the normal operation of the pager; the feature classification module is used to use the rbf neural network to decide the objects that the user needs to call; the storage module is used to store user settings information; the 4G network module is used for communication with the object called by the user.

The wireless headset is used to collect the user's EEG information and send it to the pager through Bluetooth to provide the microprocessor inside the pager for further processing; the microprocessor is used to parse the received EEG signal and send it to the communication device of the called party.

The training method of the automatic pager of the above-mentioned brain-computer interface, as shown in Figure 3, includes the following steps:

Step 1: Determine the character characteristic value of the tested person and set it on the display screen of the pager, and set the character characteristic value as x ₄ ;

Step 2: Put on the wireless headset and turn on the switch;

Step 3: Use the wireless headset to transmit the tester's EEG signal in a specific environment to the pager, and the feature extraction module of the pager will extract the tester's EEG signal feature value in this environment, that is, the EEG signal of the tester. The maximum peak value, the minimum peak value, and the time value of the EEG signal fluctuate greatly, and the wireless headset switch is turned off at the same time, and the maximum peak value, the minimum peak value, and the large fluctuation time value of the EEG signal are set to x ₁ , x ₂ , x ₃ ;

Step 4: Determine the event characteristic value corresponding to this specific environment, and set the event characteristic value as y _d ;

Step 5: The feature value of the event called by the pager is used as the ideal output data of the feature classification module. The maximum peak value, the minimum peak value, the time of EEG signal fluctuation, and the user character feature value are the input data. Use these data to analyze the rbf neural network of the feature module. network for training;

Step 6: Start over from step 1 and replace the rbf neural network training of the feature classification module by different testers.

Embodiment 2: The pager needs to store the information of the call object and event information that may be used before the user uses it in the pager according to the corresponding prompts on the display screen. Characteristic The setting of the characteristic value of the character and the setting of the characteristic value of the calling object are carried out under the corresponding prompts on the display screen, as shown in Figure 1. The information is stored under the guidance of the relevant technical personnel, and the user can make a call as long as he sets the character characteristic value and the called object according to his own character.

The working principle of the present invention: before the user uses the pager, first set the object of the call according to his own needs and set his own character characteristic value x ₄ and the storage of the event information and the information of the called object. The storage of this information can be stored in Refer to Tables 1 and 2 to complete under the guidance of relevant technical personnel.

Table 1 Call object characteristic value table

Called object characteristic value Callee's name Contact information of the person called 0 Father *** 1 Mother *** 2 elder brother *** 3 elder sister *** 4 younger brother *** 5 younger sister *** … … …

Table 2 Event Information Table

The feature value of the event information needs to be consistent with the event feature value of the pager training; then, the user puts on the wireless headset and turns on the switch. When a call is required in a certain situation, the wireless headset will convert the user's EEG signal in this situation. Collect, send to the pager through the bluetooth device, and extract the eigenvalues of the EEG signal in the feature extraction module of the microprocessor, namely x ₁ , x ₂ , x ₃ , according to the rbf neural network that has been trained in the feature classification module. The event characteristic value under the current situation can be obtained, as shown in formula (4):

y is the event characteristic value under the current situation, the control unit will send the information corresponding to the situation characteristic value in Table 2 to the call object communication device preset by the user through the 4G module, and the user uses the trained pager schematic diagram as shown in Fig. shown in Figure 2.

Embodiment 3: The use of the pager is described by taking an example when property is threatened. At this time, it is assumed that the user's property is threatened and needs to call for help. The user puts on the wireless headset and turns on the switch, and the microprocessor of the pager performs feature extraction on the received EEG signals. The trained rbf in the feature classification module is based on the current The input data predicts that the event characteristic value of the current situation is 2, and the microprocessor of the pager will effectively transmit the information to the communication device of the called object according to the stored information.

When in use, when the user needs to call in a specific environment, the user only needs to put on the wireless headset and turn on the switch for 3 to 5 seconds, and the pager can automatically send the call information to the callee.

In the above-mentioned use process, there are few steps that the user needs to operate by himself. When inputting the information of the called object, the user can input the information by himself or let the technical personnel who provide the product complete it. For the information input of the called object, refer to Table 1 for operations. The microprocessor can use ARM series processors.

Although the present invention has been described in more detail above with reference to the accompanying drawings and accompanying tables, it is not limited, so the present invention is not limited to the above embodiments. As long as it is based on the present invention and no other time-consuming and laborious work is performed, it is covered within the scope of the claims of the present invention.

Claims (5)

1. A training method of an automatic calling device based on a brain-computer interface is characterized in that the automatic calling device comprises a wireless earphone and a calling device, a microprocessor is arranged in the calling device, the microprocessor is respectively connected with a Bluetooth receiving device, a preprocessing module, a feature extraction module, a feature classification module, a 4G network module, a power supply module and a storage module, a display screen and button keys are arranged on the surface of the calling device, corresponding operations can be completed on the display screen according to corresponding prompts, and the button keys are connected with the internal microprocessor; the Bluetooth receiving device is used for receiving the electroencephalogram signals sent by the wireless earphone; the preprocessing module is used for denoising and signal amplification of the received electroencephalogram signals; the characteristic extraction module is used for analyzing and extracting the maximum peak value and the minimum peak value of the received electroencephalogram signal and the large fluctuation time of the electroencephalogram signal; the power supply module is used for providing electric energy for the normal operation of the caller; the characteristic classification module is used for deciding the object to be called by the user by using an RBF algorithm; the storage module is used for storing information set by a user; the 4G network module is used for communicating with an object called by a user;

the wireless earphone is used for acquiring electroencephalogram information of a user and sending the electroencephalogram information to the caller through Bluetooth so as to provide the electroencephalogram information for the microprocessor in the caller for further processing; the microprocessor is used for analyzing the received brain electrical signal and transmitting the brain electrical signal to the communication equipment of the called object;

the training method of the automatic caller comprises the following steps:

step 1: determining character characteristic value of the tested person and setting the character characteristic value on the display screen of the pager, wherein the character characteristic value is x₄；

Step 2: wearing a wireless earphone and turning on a switch;

and step 3: the wireless earphone is utilized to transmit the EEG signal of a tester in a certain specific environment to the caller, and the feature extraction module of the caller extracts the EEG signal feature value of the tester in the environment, namely the EEG signalThe maximum peak value, the minimum peak value and the electroencephalogram signal large fluctuation time value are simultaneously closed, and the maximum peak value, the minimum peak value and the electroencephalogram signal large fluctuation time value are set as x₁、x₂、x₃；

And 4, step 4: determining the corresponding event characteristic value under the specific environment, and setting the event characteristic value as y_d；

And 5: the characteristic value of an event called by a caller is used as ideal output data of a characteristic classification module, the maximum peak value, the minimum peak value, the large fluctuation time of an electroencephalogram signal and the characteristic value of the user character are used as input data, and the data are utilized to train the rbf neural network of the characteristic module;

step 6: and (3) restarting to replace the training of the rbf neural network of the feature classification module by different testers from the step 1.

2. The method for training an automatic pager based on brain-computer interface as claimed in claim 1, wherein: the setting of the character characteristic value and the setting of the event characteristic value are finished through the prompt of a display screen of a calling device, the setting of the characteristic value of the event is only used for training the calling device by technicians, and common users cannot use the setting.

3. The method for training an automatic pager based on brain-computer interface as claimed in claim 1, wherein: the specific environments are divided into three categories: the system comprises a general case, an emergency case and a super-emergency case, wherein the general case takes an event characteristic value as 1, the emergency case takes an event characteristic value as 2, the super-emergency case takes an event characteristic value as 3, and an event characteristic value table can be stored in a storage unit in advance, wherein the general case is defined as trivial matters needing help in daily life; an emergency is defined as a situation where property or interest is compromised; a super-emergency is defined as a situation where life safety is compromised.

4. The method for training an automatic pager based on brain-computer interface as claimed in claim 1, wherein: the character characteristic values comprise 0, 1 and 2, wherein 0 is a character which represents inward comparison, 1 represents a character which is neither inward nor outward, and 2 represents outward character, and the character characteristic values need to be evaluated by a user according to self-understanding and people around the user in daily life.

5. The method for training an automatic pager based on brain-computer interface as claimed in claim 1, wherein: the training of the rbf neural network of the feature classification module comprises the following specific steps: the training algorithm has four inputs, a hidden layer and an output, the hidden layer uses a Gaussian function as a radial basis function, and the calculation formula of the output of the first hidden layer is as follows:

wherein i ∈ [1, 10 ]]10 neurons representing the first hidden layer, X_cRepresenting the central value of a Gaussian function, which is a column vector, σ, co-dimensional with X_iIs the width of the gaussian function, X is the set of input vectors,

the output of the network is thus:

wherein sigma_i、X_cAnd ω_i1Randomly giving to calculate the actual output y of the first group of input data₁Then update σ according to a gradient descent algorithm_i、X_c、ω_i1Continuing the second group until all the input and output data are trained on the rbf neural network to obtain the final sigma_i、X_c、ω_i1And (4) finishing.

Images