CN114038450A - Dialect identification method, dialect identification device, dialect identification equipment and storage medium - Google Patents

Abstract

The invention relates to an artificial intelligence technology, and discloses a dialect identification method, which comprises the following steps: receiving dialect voice data input by a user, and extracting voice characteristics of the dialect voice data; carrying out similarity detection on the speech features one by utilizing training data corresponding to all dialect recognition models in a pre-constructed dialect model library to obtain similarity scores of the speech features and each type of training data; taking the dialect recognition model corresponding to the training data with the highest similarity score as a target dialect recognition model; and converting the dialect voice data by using the target dialect recognition model to obtain a voice recognition result corresponding to the dialect voice data. The invention also provides a dialect identification device, an electronic device and a storage medium. The invention can solve the problem of low dialect identification precision.

Description

Dialect identification method, device, device and storage medium

technical field

The present invention relates to the technical field of artificial intelligence, and in particular, to a dialect identification method, apparatus, electronic device, and computer-readable storage medium.

Background technique

With the development of society, more and more software, such as input method, navigation software, intelligent question answering system, etc., must use language recognition technology. Speech recognition technology is gradually becoming the key technology of human-computer interaction in information technology. At present, speech recognition mostly recognizes Mandarin, and dialect, as a local language, is still used by a large number of people, especially some older people who do not speak Mandarin but only speak dialect. Therefore, the speech recognition of dialect is also an important research topic.

At present, dialect recognition is mostly performed by dialect training on a model, and then the trained model is used to recognize the dialect. However, the types of dialects vary widely, and each dialect recognition model may only recognize one or several dialects. Therefore, it is usually necessary to Many dialect recognition models are trained, and the multi-dialect recognition models are brought together to form a multi-dialect recognition model. Choose a suitable dialect recognition model, so that the accuracy of dialect recognition is not high.

SUMMARY OF THE INVENTION

The present invention provides a dialect identification method, device and computer-readable storage medium, the main purpose of which is to solve the problem of low dialect identification accuracy.

To achieve the above object, a dialect identification method provided by the present invention includes:

receiving the dialect voice data input by the user, and extracting the voice features of the dialect voice data;

Utilize the training data corresponding to all dialect recognition models in the pre-built dialect model library to perform similarity detection with the voice features one by one, and obtain the similarity score between the voice feature and each of the training data;

Using the dialect recognition model corresponding to the training data with the highest similarity score as the target dialect recognition model;

The dialect speech data is converted by using the target dialect recognition model to obtain a speech recognition result corresponding to the dialect speech data.

Optionally, the extracting the voice features of the dialect voice data includes:

converting the sound signal in the voice data into a digital signal;

The digital signal is calculated by using a preset triangular band-pass filter to obtain the speech feature corresponding to the speech data.

Optionally, the digital signal is calculated by using a preset triangular bandpass filter to obtain the speech feature corresponding to the speech data, including:

Performing pre-emphasis, framing and windowing processing on the digital signal to obtain frequency domain energy;

performing fast Fourier transform on the frequency domain energy to obtain a frequency spectrum;

Using the triangular bandpass filter to calculate the spectrum to obtain logarithmic energy;

Perform discrete cosine transform on the logarithmic energy to obtain Mel frequency cepstral coefficients;

The difference calculation is performed according to the Mel frequency cepstral coefficients to obtain a dynamic difference parameter, and the dynamic difference parameter is determined as a speech feature.

Optionally, the training data corresponding to all dialect recognition models in the pre-built dialect model library is used to perform similarity detection with the voice features one by one, to obtain the similarity score between the voice feature and each of the training data. values, including:

Extracting the phonetic features of the training data of each of the dialect recognition models one by one;

Calculate the distance value of the voice feature of the training data and the voice feature of the dialect voice data input by the user;

The similarity score between the voice feature of the dialect voice data input by the user and the voice feature of each type of the training data is calculated according to the distance value.

Optionally, the use of the target dialect recognition model to convert the dialect voice data to obtain a voice recognition result corresponding to the dialect voice data, including:

Using the target dialect recognition model to perform a preset number of convolution, pooling and full connection operations on the dialect voice data to obtain a coding vector;

The encoding vector is decoded using a preset activation function to obtain a speech recognition result.

Optionally, before the dialect recognition model corresponding to the training data with the highest similarity score is used as the target dialect recognition model, the method further includes:

Obtain random voice training sets and multi-dialect voice training sets;

Trained from the random speech training set using a pre-built general speech model. get the general speech model;

According to the general speech model and the multi-dialect speech training set, adaptive training is performed respectively to obtain a plurality of dialect recognition models.

In order to solve the above problems, the present invention also provides a dialect identification device, the device includes:

A voice feature extraction module for receiving dialect voice data input by a user, and extracting voice features of the dialect voice data;

The target dialect recognition model determination module is used to perform similarity detection with the voice features one by one using the training data corresponding to all dialect recognition models in the pre-built dialect model library, and obtain the relationship between the voice features and each of the training data. similarity score; take the dialect recognition model corresponding to the training data with the highest similarity score as the target dialect recognition model;

The speech recognition result generation module is used for converting the dialect speech data by using the target dialect recognition model to obtain a speech recognition result corresponding to the dialect speech data.

In order to solve the above problems, the present invention also provides an electronic device, the electronic device includes:

at least one processor; and,

a memory communicatively coupled to the at least one processor; wherein,

The memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to execute the dialect recognition method described above.

In order to solve the above problems, the present invention also provides a computer-readable storage medium, where at least one computer program is stored in the computer-readable storage medium, and the at least one computer program is executed by a processor in an electronic device to realize the above-mentioned The dialect identification method described above.

In the embodiment of the present invention, the corresponding dialect recognition model is obtained by using the speech features of the dialect voice data and the training data of the dialect recognition model in the dialect model database. Each dialect recognition model in the dialect model database corresponds to this dialect. The corresponding dialect recognition model is used to recognize the speech data, which improves the recognition accuracy; the speech recognition result corresponding to the dialect speech data is obtained through the dialect recognition model, and the accuracy of the speech recognition result obtained by the dialect recognition is improved. Therefore, the dialect identification method, device, electronic device and computer-readable storage medium proposed by the present invention can solve the problem of low dialect identification accuracy.

Description of drawings

1 is a schematic flowchart of a dialect identification method provided by an embodiment of the present invention;

FIG. 2 is a schematic flowchart of obtaining a similarity score according to an embodiment of the present invention;

3 is a schematic flowchart of obtaining a speech recognition result according to an embodiment of the present invention;

4 is a functional block diagram of a dialect identification device provided by an embodiment of the present invention;

FIG. 5 is a schematic structural diagram of an electronic device implementing the dialect identification method according to an embodiment of the present invention.

The realization, functional characteristics and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments.

Detailed ways

It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention.

The embodiment of the present application provides a dialect identification method. The execution subject of the dialect identification method includes, but is not limited to, at least one of electronic devices that can be configured to execute the method provided by the embodiments of the present application, such as a server and a terminal. In other words, the dialect identification method can be executed by software or hardware installed in a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, and the like. The server can be an independent server, or can provide cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery network (Content Delivery Network) Network, CDN), and cloud servers for basic cloud computing services such as big data and artificial intelligence platforms.

Referring to FIG. 1 , it is a schematic flowchart of a dialect identification method provided by an embodiment of the present invention.

In this embodiment, the dialect identification method includes:

S1, receive the dialect voice data input by the user, and extract the voice feature of the dialect voice data;

In this embodiment of the present invention, the dialect voice data may be any dialect. The dialects are commonly known as local dialects, which are only used in certain regions, and are not another language independent of Chinese, but only languages used in local areas, such as Cantonese, Hokkien, Hakka and so on.

In this embodiment of the present invention, the extraction of the voice features of the dialect voice data includes:

converting the sound signal in the voice data into a digital signal;

The digital signal is calculated by using a preset triangular band-pass filter to obtain the speech feature corresponding to the speech data.

In this embodiment of the present invention, the sound signal in the voice data may be converted into a digital signal through steps such as sampling, quantization, and encoding.

In detail, the sampling refers to the amplitude of the sound signal obtained at a specific moment; through sampling, a time-continuous analog signal (voice data) can be converted into a time-discrete, amplitude-continuous discrete signal.

In one embodiment of the present invention, the language information may be sampled according to a fixed time interval as a sampling period. In the quantization step, each sample that is continuously valued in magnitude is converted to a discrete-valued representation.

Further, the digital signal is calculated by using a preset triangular bandpass filter to obtain the speech feature corresponding to the speech data, including:

Performing pre-emphasis, framing and windowing processing on the digital signal to obtain frequency domain energy;

performing fast Fourier transform on the frequency domain energy to obtain a frequency spectrum;

Using the triangular bandpass filter to calculate the spectrum to obtain logarithmic energy;

Perform discrete cosine transform on the logarithmic energy to obtain Mel frequency cepstral coefficients;

The difference calculation is performed according to the Mel frequency cepstral coefficients to obtain a dynamic difference parameter, and the dynamic difference parameter is determined as a speech feature.

In the embodiment of the present invention, the pre-emphasis refers to passing the digital signal through a high-pass filter to enhance the high-frequency part, so that the frequency spectrum of the signal becomes flat and kept in the entire frequency band from low frequency to high frequency. The framing is to divide the data in the digital signal by taking the data collected in a preset unit time as one frame. The windowing is to multiply the digital signal of each frame by a Hamming window, which can increase the continuity of the left end and the right end of a frame. By pre-emphasizing, framing and windowing the digital signal, the effects of vocal cords and lips during the vocalization process can be eliminated, the high-frequency part of the speech signal that is suppressed by the vocalization system can be compensated, and the windowed digital signal can be converted into frequency The energy distribution on the domain, and different energy distributions can represent the characteristics of different speech.

The triangular band-pass filter can reduce the amount of computation and has the effect of eliminating harmonics by smoothing the frequency spectrum, so as to highlight the formant of the speech. Therefore, the pitch or pitch of a speech will not appear in the Mel-frequency cepstral coefficients, so the Mel-frequency cepstral coefficients will not be affected by the pitch of the input speech. The standard Mel-frequency cepstral coefficients Only the static characteristics of speech are reflected, and the dynamic characteristics of speech can be described by the difference spectrum of the static characteristics. The dynamic difference parameter is a combination of dynamic and static characteristics, which can effectively improve the recognition performance of the system.

The standard Mel-frequency cepstral coefficients only reflect the static characteristics of speech, and the dynamic characteristics of speech can be described by the differential spectrum of the static features. Voice characteristics.

S2, utilize the training data corresponding to all dialect recognition models in the pre-built dialect model library to perform similarity detection with the voice feature one by one, to obtain the similarity score between the voice feature and each of the training data;

In the embodiment of the present invention, each dialect recognition model in the pre-built dialect model library can be constructed by a convolutional neural network. After training through different dialect voice training sets, the corresponding dialect voice can be converted into text data. For example, the dialect recognition model A is obtained by training Cantonese, and the trained dialect recognition model A can convert Cantonese into text data. Wherein, each dialect voice training set includes a dialect label, and the dialect label is used to identify the dialect type.

In the embodiment of the present invention, referring to FIG. 2 , the training data corresponding to all dialect recognition models in the pre-built dialect model library is used to perform similarity detection with the voice features one by one, and the voice features and each type of voice feature are obtained. The similarity score of the training data, including:

S21, extract the speech features of the training data of each of the dialect recognition models one by one;

S22, calculate the distance value of the voice feature of the training data and the voice feature of the dialect voice data input by the user;

S23. Calculate, according to the distance value, a similarity score between the voice feature of the dialect voice data input by the user and the voice feature of each type of the training data.

Further, in this embodiment of the present invention, the distance value between the voice feature of the training data and the voice feature of the dialect voice data input by the user can be calculated by the following formula:

Wherein, D is the distance value, R _i is the voice feature of the training data of the dialect recognition model, T is the voice feature of the dialect voice data input by the user, and θ is a preset coefficient.

通过公式计算得到所述语音特征A与所述语音特征

的距离值为40，本发明实施例根据预设规则计算，例如：相似度分值为1-距离值/100，则所述方言识别模型的训练数据语音特征的相似度分值为0.6。For example, the voice feature of the training data of the dialect recognition model is A, and the voice feature of the dialect voice data input by the user is

The voice feature A and the voice feature are obtained by formula calculation

The distance value is 40, which is calculated according to a preset rule in this embodiment of the present invention. For example, if the similarity score is 1-distance value/100, the similarity score of the speech feature of the training data of the dialect recognition model is 0.6.

S3, using the dialect recognition model corresponding to the training data with the highest similarity score as the target dialect recognition model;

In the embodiment of the present invention, a higher similarity score indicates that the voice feature is more similar. Therefore, in the embodiment of the present invention, the dialect recognition model with the highest score is used as the target dialect recognition model of the dialect voice data.

In one embodiment of the present invention, the dialect label of the training data with the highest similarity score can be extracted as the dialect label of the dialect voice data, and the dialect label can let the user know its voice data through front-end display or other means Identify the type of dialect obtained.

S4. Convert the dialect voice data by using the target dialect recognition model to obtain a voice recognition result corresponding to the dialect voice data.

Referring to FIG. 3 , in this embodiment of the present invention, the S4 includes:

S41, using the target dialect recognition model to perform a preset number of convolution, pooling and full connection operations on the dialect voice data to obtain a coding vector;

S42. Decode the encoding vector by using a preset activation function to obtain a speech recognition result.

In this embodiment of the present invention, the speech recognition result may be text data, and in this embodiment of the present invention, a CNN network or an RNN network may be used to perform convolution, pooling, and full connection operations on the dialect speech data. For example, using the RNN network to perform convolution, pooling and full connection operations on the dialect speech data to obtain a coding vector; and then use a single-layer neural network with a classifier as a decoding layer to decode the coding vector. The decoding layer adopts activation functions such as softmax activation function and sigmoid activation function.

In another embodiment of the present invention, before the dialect recognition model corresponding to the training data with the highest similarity score is used as the target dialect recognition model, the method further includes:

Obtain random voice training sets and multi-dialect voice training sets;

Utilize pre-built general speech model to train according to described random speech training set, obtain general speech model;

According to the general speech model and the multi-dialect speech training set, adaptive training is performed respectively to obtain a plurality of dialect recognition models.

In the embodiment of the present invention, the pre-built universal voice model (UBM, Universal Background Model) is a universal voice model obtained by first collecting a large number of random voices and training, and then using part of the dialect voice data to adjust the universal voice through an adaptive algorithm The parameters of the model are obtained to obtain the dialect recognition model.

In the embodiment of the present invention, the voice features of the dialect voice A are adapted through the general voice model, so that the dialect recognition model of the dialect voice A can be obtained more quickly, and too many dialect voices A are not needed as the extraction target of voice features Then the dialect recognition model can be obtained.

In this embodiment of the present invention, the speech features may be scattered around certain Gaussian distributions of the general speech model. The adaptive process is to shift each Gaussian distribution of the general speech model to the speech feature, specifically including: using the speech feature to calculate the update parameters (Gaussian weight, mean and variance) of the general speech model; The obtained updated parameters are fused with the original parameters of the general speech model, thereby obtaining a dialect recognition model suitable for the speech features. The adaptive algorithms include, but are not limited to, Maximum A posteriori (MAP), Maximum Likelihood Linear Regression (MLLR).

In the embodiment of the present invention, the general speech model is fine-tuned to the dialect recognition model of the speech feature through an adaptive algorithm. This method greatly reduces the sample size and training time required for training by reducing the training parameters.

In another embodiment of the present invention, after converting the dialect voice data by using the target dialect recognition model to obtain a voice recognition result corresponding to the dialect voice data, the method may further include: performing the voice recognition result on the voice recognition result. Text conversion to obtain the Mandarin text corresponding to the dialect voice data.

In the embodiment of the present invention, performing text conversion on the ordinary voice data obtained by converting the dialect voice data can enable the user to view the text content of the dialect voice more intuitively.

In the embodiment of the present invention, the corresponding dialect recognition model is obtained by using the speech features of the dialect voice data and the training data of the dialect recognition model in the dialect model database. Each dialect recognition model in the dialect model database corresponds to this dialect. The corresponding dialect recognition model is used to recognize the dialect voice data, which improves the recognition accuracy. Therefore, the dialect identification method proposed by the present invention can solve the problem of low dialect identification accuracy.

As shown in FIG. 4 , it is a functional block diagram of a dialect identification device provided by an embodiment of the present invention.

The dialect identification apparatus 100 of the present invention can be installed in an electronic device. According to the implemented functions, the dialect recognition apparatus 100 may include a speech feature extraction module 101 , a target dialect recognition model determination module 102 and a speech recognition result generation module 103 . The modules in the present invention can also be called units, which refer to a series of computer program segments that can be executed by the electronic device processor and can perform fixed functions, and are stored in the memory of the electronic device.

In this embodiment, the functions of each module/unit are as follows:

The voice feature extraction module 101 is configured to receive the dialect voice data input by the user, and extract the voice features of the dialect voice data;

The target dialect recognition model determination module 102 is used to perform similarity detection with the voice features one by one using the training data corresponding to all dialect recognition models in the pre-built dialect model library, and obtain the voice features and each of the described voice features. The similarity score of the training data; the dialect recognition model corresponding to the training data with the highest similarity score is used as the target dialect recognition model;

The speech recognition result generating module 103 is configured to convert the dialect speech data by using the target dialect recognition model to obtain a speech recognition result corresponding to the dialect speech data.

In detail, each module described in the dialect identification device 100 in the embodiment of the present invention adopts the same technical means as the dialect identification method described in the above-mentioned FIG. 1 to FIG. 3 during use, and can produce the same technical effect , which will not be repeated here.

As shown in FIG. 5 , it is a schematic structural diagram of an electronic device for implementing a dialect identification method provided by an embodiment of the present invention.

The electronic device 1 may include a processor 10, a memory 11, a communication bus 12, and a communication interface 13, and may also include a computer program stored in the memory 11 and executable on the processor 10, such as a dialect recognition program .

The processor 10 may be composed of integrated circuits in some embodiments, for example, may be composed of a single packaged integrated circuit, or may be composed of multiple integrated circuits packaged with the same function or different functions, including one or A combination of multiple central processing units (Central Processing Units, CPUs), microprocessors, digital processing chips, graphics processors, and various control chips, etc. The processor 10 is the control core (ControlUnit) of the electronic device, and uses various interfaces and lines to connect the various components of the entire electronic device, by running or executing the program or module (for example, executing the dialect) stored in the memory 11. identification programs, etc.), and call data stored in the memory 11 to perform various functions of the electronic device and process data.

The memory 11 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (for example: SD or DX memory, etc.), magnetic memory, magnetic disk, optical disk, etc. . In some embodiments, the memory 11 may be an internal storage unit of an electronic device, such as a mobile hard disk of the electronic device. In other embodiments, the memory 11 may also be an external storage device of the electronic device, such as a pluggable mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) equipped on the electronic device. ) card, flash card (Flash Card) and so on. Further, the memory 11 may also include both an internal storage unit of an electronic device and an external storage device. The memory 11 can not only be used to store application software installed in the electronic device and various types of data, such as codes of dialect recognition programs, etc., but also can be used to temporarily store data that has been output or will be output.

The communication bus 12 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus or the like. The bus can be divided into address bus, data bus, control bus and so on. The bus is configured to implement connection communication between the memory 11 and at least one processor 10 and the like.

The communication interface 13 is used for communication between the above electronic device and other devices, including a network interface and a user interface. Optionally, the network interface may include a wired interface and/or a wireless interface (eg, a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device and other electronic devices. The user interface may be a display (Display), an input unit (such as a keyboard (Keyboard)), and optionally, the user interface may also be a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode, organic light-emitting diode) touch device, and the like. The display may also be appropriately called a display screen or a display unit, which is used for displaying information processed in the electronic device and for displaying a visual user interface.

FIG. 5 only shows an electronic device with components. Those skilled in the art can understand that the structure shown in FIG. 5 does not constitute a limitation on the electronic device 1, and may include fewer or more components than those shown in the drawings. components, or a combination of certain components, or a different arrangement of components.

For example, although not shown, the electronic device may also include a power source (such as a battery) for powering the various components, preferably, the power source may be logically connected to the at least one processor 10 through a power management device, so that the power source can be logically connected through the power management device. Implement functions such as charge management, discharge management, and power management. The power source may also include one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, and any other components. The electronic device may further include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

It should be understood that the embodiments are only used for illustration, and are not limited by this structure in the scope of the patent application.

The dialect recognition program stored in the memory 11 in the electronic device 1 is a combination of multiple instructions, and when running in the processor 10, it can realize:

receiving the dialect voice data input by the user, and extracting the voice features of the dialect voice data;

Utilize the training data corresponding to all dialect recognition models in the pre-built dialect model library to perform similarity detection with the voice features one by one, and obtain the similarity score between the voice feature and each of the training data;

Using the dialect recognition model corresponding to the training data with the highest similarity score as the target dialect recognition model;

The dialect speech data is converted by using the target dialect recognition model to obtain a speech recognition result corresponding to the dialect speech data.

Specifically, for the specific implementation method of the above-mentioned instruction by the processor 10, reference may be made to the description of the relevant steps in the embodiment corresponding to the accompanying drawings, which will not be repeated here.

Further, if the modules/units integrated in the electronic device 1 are implemented in the form of software functional units and sold or used as independent products, they may be stored in a computer-readable storage medium. The computer-readable storage medium may be volatile or non-volatile. For example, the computer-readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a removable hard disk, a magnetic disk, an optical disc, a computer memory, a read-only memory (ROM, Read-Only). Memory).

The present invention also provides a computer-readable storage medium, where the readable storage medium stores a computer program, and when executed by a processor of an electronic device, the computer program can realize:

receiving the dialect voice data input by the user, and extracting the voice features of the dialect voice data;

Utilize the training data corresponding to all dialect recognition models in the pre-built dialect model library to perform similarity detection with the voice features one by one, and obtain the similarity score between the voice feature and each of the training data;

Using the dialect recognition model corresponding to the training data with the highest similarity score as the target dialect recognition model;

The dialect speech data is converted by using the target dialect recognition model to obtain a speech recognition result corresponding to the dialect speech data.

In the several embodiments provided by the present invention, it should be understood that the disclosed apparatus, apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are only illustrative. For example, the division of the modules is only a logical function division, and there may be other division manners in actual implementation.

The modules described as separate components may or may not be physically separated, and components shown as modules may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution in this embodiment.

In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware, or may be implemented in the form of hardware plus software function modules.

It will be apparent to those skilled in the art that the present invention is not limited to the details of the above-described exemplary embodiments, but that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics of the invention.

Therefore, the embodiments are to be regarded in all respects as illustrative and not restrictive, and the scope of the invention is defined by the appended claims rather than the foregoing description, which are therefore intended to fall within the scope of the appended claims. All changes within the meaning and range of the equivalents of , are included in the present invention. Any reference signs in the claims shall not be construed as limiting the involved claim.

The blockchain referred to in the present invention is a new application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm. Blockchain, essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information to verify its Validity of information (anti-counterfeiting) and generation of the next block. The blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.

The embodiments of the present application may acquire and process related data based on artificial intelligence technology. Among them, artificial intelligence (AI) is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. .

Furthermore, it is clear that the word "comprising" does not exclude other units or steps and the singular does not exclude the plural. Several units or means recited in the system claims can also be implemented by one unit or means by means of software or hardware. The words first, second, etc. are used to denote names and do not denote any particular order.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be Modifications or equivalent substitutions can be made without departing from the spirit and scope of the technical solutions of the present invention.

Claims (9)

1. A dialect identification method, the method comprising:

receiving dialect voice data input by a user, and extracting voice characteristics of the dialect voice data;

carrying out similarity detection on the speech features one by utilizing training data corresponding to all dialect recognition models in a pre-constructed dialect model library to obtain similarity scores of the speech features and each type of training data;

taking the dialect recognition model corresponding to the training data with the highest similarity score as a target dialect recognition model;

and converting the dialect voice data by using the target dialect recognition model to obtain a voice recognition result corresponding to the dialect voice data.

2. The dialect recognition method of claim 1, wherein said extracting speech features of the dialect speech data comprises:

converting the sound signals in the voice data into digital signals;

and calculating the digital signal by using a preset triangular band-pass filter to obtain the voice characteristics corresponding to the voice data.

3. The dialect recognition method of claim 2, wherein the calculating the digital signal by using a preset triangular band-pass filter to obtain the voice feature corresponding to the voice data comprises:

pre-emphasis, framing and windowing are carried out on the digital signal to obtain frequency domain energy;

performing fast Fourier transform on the frequency domain energy to obtain a frequency spectrum;

calculating the frequency spectrum by using the triangular band-pass filter to obtain logarithmic energy;

discrete cosine transform is carried out on the logarithmic energy to obtain a Mel frequency cepstrum coefficient;

and carrying out differential calculation according to the mel frequency cepstrum coefficient to obtain a dynamic differential parameter, and determining the dynamic differential parameter as a voice characteristic.

4. The dialect recognition method of claim 1, wherein the performing similarity detection with the speech features one by using training data corresponding to all dialect recognition models in a pre-constructed dialect model library to obtain similarity scores of the speech features and each kind of the training data comprises:

extracting the voice characteristics of the training data of each dialect recognition model one by one;

calculating a distance value between the voice feature of the training data and the voice feature of the dialect voice data input by the user;

and calculating the similarity score of the voice feature of the dialect voice data input by the user and the voice feature of each type of training data according to the distance value.

5. The dialect recognition method of claim 1, wherein the converting the dialect speech data using the target dialect recognition model to obtain the speech recognition result corresponding to the dialect speech data comprises:

carrying out convolution, pooling and full-connection operation on the dialect voice data for preset times by using the target dialect identification model to obtain a coding vector;

and decoding the coding vector by using a preset activation function to obtain a voice recognition result.

6. The dialect recognition method of any one of claims 1 to 5, wherein before the dialect recognition model corresponding to the training data with the highest similarity score is taken as the target dialect recognition model, the method further comprises:

acquiring a random voice training set and a multi-party speech sound training set;

training according to the random voice training set by using a pre-constructed general voice model to obtain a general voice model;

and respectively carrying out self-adaptive training according to the general voice model and the multi-dialect voice training set to obtain a plurality of dialect recognition models.

7. A dialect recognition apparatus, the apparatus comprising:

the voice feature extraction module is used for receiving dialect voice data input by a user and extracting voice features of the dialect voice data;

the target dialect recognition model determining module is used for carrying out similarity detection on the training data corresponding to all the dialect recognition models in the pre-constructed dialect model library and the voice features one by one to obtain similarity scores of the voice features and each type of the training data; taking the dialect recognition model corresponding to the training data with the highest similarity score as a target dialect recognition model;

and the voice recognition result generation module is used for converting the dialect voice data by using the target dialect recognition model to obtain a voice recognition result corresponding to the dialect voice data.

8. An electronic device, characterized in that the electronic device comprises:

at least one processor; and the number of the first and second groups,

a memory communicatively coupled to the at least one processor; wherein,

the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the dialect identification method of any one of claims 1 to 7.

9. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out a dialect identification method according to any one of claims 1 to 7.

Images