Speech recognition method and device based on brain-like model, electronic equipment and storage medium

Abstract

The invention provides a voice recognition method and device based on a brain-like model, electronic equipment and a storage medium, and the method comprises the steps: obtaining a whole-brain network topological structure according to a brain function network generated by human brain image data, and carrying out the recognition of a whole-brain network through employing a multi-class neuron model as a node and a synaptic plasticity model as an edge, constructing a multi-brain-region pulse neural network as a brain-like model; constructing a speech recognition framework of the brain-like model; electromagnetic intervention is applied to different brain areas of the brain-like model, optimal electromagnetic intervention parameters are determined by analyzing the voice recognition accuracy of the brain-like model before and after electromagnetic intervention, and brain-like model voice recognition is carried out according to the optimal electromagnetic intervention parameters. According to the invention, the speech recognition performance of the brain-like model can be effectively improved, the biological interpretability and neural information processing capability of the brain-like model are further improved, and the development of brain-like intelligence in the application of a mode recognition task is promoted.

Description

Technical Field

[0001] The present application belongs to the field of brain science research technology, and in particular relates to a speech recognition method, device, electronic device and storage medium based on a brain-like model. Background Art

[0002] Pattern recognition, as a fundamental ability of human cognition, is the cornerstone of the brain's higher-level cognitive functions. At the same time, pattern recognition is also one of the core goals of artificial intelligence, and many complex intelligent problems often rely on automatic and accurate pattern recognition. Brain-inspired models based on spiking neural networks simulate the spiking discharge behavior of neurons, which is more consistent with biological neural information processing mechanisms. Therefore, they have advantages in handling speech recognition tasks that contain time-series related information. Therefore, brain-inspired models have attracted increasing attention from scholars in speech recognition tasks. However, the recognition performance of existing brain-inspired models in complex environments and tasks needs to be further improved. Further exploration and development of brain-inspired models in the field of pattern recognition is expected to further enhance the pattern recognition capabilities of brain-inspired models and promote the development of brain-inspired intelligence. Summary of the Invention

[0003] In view of this, the present application aims to propose a speech recognition method, device, electronic device and storage medium based on a brain-inspired model to solve the problem of insufficient pattern recognition performance of the current brain-inspired model.

[0004] To achieve the above objectives, the technical solution of this application is implemented as follows:

[0005] In a first aspect, the present application provides a speech recognition method based on a brain-inspired model, comprising:

[0006] The brain functional network generated from human brain imaging data was used to obtain the whole-brain network topology. A multi-brain region spiking neural network was constructed as a brain-like model using multiple types of neuron models as nodes and synaptic plasticity models as edges.

[0007] Constructing a speech recognition framework based on a brain-inspired model, wherein the speech recognition framework includes input layer neurons, a reservoir layer model, and output layer neurons, takes a speech pulse sequence as input, and trains the synaptic weights between the reservoir layer model and the output layer neurons using a remote supervised learning algorithm;

[0008] Electromagnetic intervention is applied to different brain regions of the brain-like model, and optimal electromagnetic intervention parameters are determined by analyzing the speech recognition accuracy of the brain-like model before and after the electromagnetic intervention, and brain-like model speech recognition is performed according to the optimal electromagnetic intervention parameters.

[0009] In a second aspect, based on the same inventive concept, the present application further provides a speech recognition device based on a brain-inspired model, comprising:

[0010] A model building module is configured to obtain a whole-brain network topology structure based on a brain functional network generated from human brain imaging data, and to construct a multi-brain region spiking neural network as a brain-like model using multiple types of neuron models as nodes and synaptic plasticity models as edges;

[0011] a speech recognition framework construction module configured to construct a speech recognition framework of a brain-inspired model, wherein the speech recognition framework includes input layer neurons, a reservoir layer model, and output layer neurons, takes a speech pulse sequence as input, and trains synaptic weights between the reservoir layer model and the output layer neurons using a remote supervised learning algorithm;

[0012] The electromagnetic intervention module is configured to apply electromagnetic intervention to different brain regions of the brain-like model, determine the optimal electromagnetic intervention parameters by analyzing the speech recognition accuracy of the brain-like model before and after the electromagnetic intervention, and perform brain-like model speech recognition based on the optimal electromagnetic intervention parameters.

[0013] In a third aspect, based on the same inventive concept, the present application also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in the first aspect when executing the program.

[0014] In a fourth aspect, based on the same inventive concept, the present application also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the method described in the first aspect.

[0015] Compared with the prior art, the speech recognition method, device, electronic device, and storage medium based on the brain-inspired model described in this application have the following beneficial effects:

[0016] The brain-like model-based speech recognition method described in this application can effectively improve the speech recognition performance of the brain-like model by applying a determined optimal electromagnetic intervention, further enhance the biological interpretability and neural information processing capabilities of the brain-like model, and promote the development of brain-like intelligence in pattern recognition tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0018] Figure 1 This is a flow chart of a speech recognition method based on a brain-inspired model described in an embodiment of the present application;

[0019] Figure 2 This is a schematic diagram of a speech recognition framework based on a brain-inspired model according to an embodiment of the present application;

[0020] Figure 3 The embodiment of this application takes 10Hz / 0.5T electromagnetic control as an example, and the applied magnetic field waveform diagram;

[0021] Figure 4 This is a schematic diagram of the recognition accuracy of the brain-like model under electromagnetic intervention described in an embodiment of the present application;

[0022] Figure 5 This is a structural diagram of a speech recognition device based on a brain-inspired model according to an embodiment of the present application;

[0023] Figure 6 This is a schematic diagram of the hardware structure of the electronic device described in an embodiment of the present application. DETAILED DESCRIPTION

[0024] In order to make the objectives, technical solutions and advantages of this application more clear, this application is further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.

[0025] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present application should have the usual meanings understood by people with ordinary skills in the field to which this application belongs. The "first", "second" and similar words used in the embodiments of the present application do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0026] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0027] Electromagnetic neuromodulation technology utilizes external electromagnetic interference to regulate nervous system function, achieving significant results in research on brain cognitive mechanisms and improving brain function. Biological research has shown that electromagnetic neuromodulation, through electromagnetic intervention in the nervous system, stimulates or inhibits brain electrical activity, thereby enhancing cognitive function and alleviating or even treating neurological and psychiatric disorders such as depression, Alzheimer's disease, and impaired consciousness. Furthermore, electromagnetic intervention can enhance interactions between brain regions and neural circuits, inducing more balanced and stable neural information transmission and functional coordination between brain regions, thereby restoring brain function to a certain extent.

[0028] Therefore, inspired by the biological effect of electromagnetic neural regulation in improving brain function, this application proposes a new method of electromagnetic intervention to improve the recognition accuracy of brain-like models, hoping to improve their performance in speech recognition tasks. This method can further enhance the neural information processing capabilities of brain-like models and promote the development of brain-like intelligence in pattern recognition tasks.

[0029] See also Figure 1 As shown, this embodiment provides a speech recognition method based on a brain-inspired model, which specifically includes the following steps:

[0030] Step S101: obtain the whole-brain network topology structure based on the brain function network generated by human brain imaging data, and construct a multi-brain region pulse neural network as a brain-like model using multiple types of neuron models as nodes and synaptic plasticity models as edges.

[0031] Specifically, the whole-brain network topology in this embodiment is as follows: based on the imaging data of the human brain (including fMRI, PET and EEG data, etc., in addition to imaging data, public data sets can also be used), a corresponding brain functional network is generated to obtain the whole-brain network topology.

[0032] Furthermore, based on the standard brain atlas, which is a digital brain structure map developed with reference to different brain anatomical structures, imaging data can be used to locate brain regions according to the brain region numbers in the atlas, dividing the brain functional network into multiple functional brain regions, thereby obtaining a multi-brain region structure of the brain-like model. Human brain imaging data is matched with the SPM standard brain atlas, and the scale of network nodes at the 90-1000 voxel level is determined according to different brain atlas templates.

[0033] The functional connectivity strength of the time series of physiological and functional metabolic measurement signals of network nodes is defined as the edge of the network. The functional connectivity strength between nodes is determined by the Pearson correlation coefficient between the average time series of physiological and functional metabolic measurement signals of different nodes. The functional connectivity matrix between nodes in each brain region is obtained by calculating the correlation coefficient between nodes in each brain region.

[0034] In order to obtain the topological structure of the brain functional network, it is necessary to set a suitable threshold X th , to determine whether there is an edge connection between nodes, different thresholds X th Different network topologies can be obtained by selecting . Characteristics of biological brain network topology: According to the characteristics of biological brain topology, the average degree of the network is greater than 2ln(N), and the network density is in the range of 3.6%-39.3%. This embodiment determines the threshold by analyzing the network topology characteristics such as network density and average node degree. The correlation coefficient between brain region nodes is greater than X th If the two nodes are connected, it is considered that there is a connection; otherwise, there is no connection. The resulting binary matrix is the final extracted network topology, which is used to construct the brain-like model.

[0035] Neuron models simulate the firing characteristics of biological neurons through mathematical modeling, including multiple general neuron models and specialized neuron models for different brain regions. Dedicated neuron models (such as hippocampal neuron models and prefrontal lobe neuron models) are assigned to the corresponding brain region nodes, while general neuron models (such as LIF neuron models, Izhikevich neuron models, and HH neuron models) are assigned to the nodes in the remaining brain regions, completing the neuron configuration of the brain-like model nodes.

[0036] Synaptic plasticity is a key mechanism for information exchange between neurons and is the foundation of learning and memory in the nervous system. The synaptic plasticity model used in this example includes both excitatory and inhibitory synapses and introduces random synaptic delays to more accurately characterize the diffusion of neurotransmitters in biological synapses. The synaptic plasticity model adjusts synaptic weights based on the timing of spike discharges between presynaptic and postsynaptic neurons to reflect the strength of the network connection, thereby enabling adaptive adjustments to input patterns.

[0037] Based on medical imaging data, the topological structure of the functional brain network was obtained. A variety of functional neuron models were introduced as nodes, and a synaptic plasticity model with coexistence of excitatory and inhibitory properties with time lag was introduced as an edge. A pulse neural network based on the topological constraints of the biological brain was constructed as a brain-like model.

[0038] Step S102: construct a speech recognition framework of a brain-like model, wherein the speech recognition framework includes input layer neurons, a reserve layer model, and output layer neurons, takes a speech pulse sequence as input, and trains the synaptic weights between the reserve layer model and the output layer neurons through a remote supervised learning algorithm.

[0039] Specifically, in this embodiment, the constructed brain-like model speech recognition framework is as follows: Figure 2 As shown in the figure, with the pulse sequence as input and the brain-like model as the reserve layer, the synaptic weights between the reserve layer and the output layer are trained by the remote supervised learning algorithm (ReSuMe) based on the Hebbian rule.

[0040] Furthermore, the speech data in this embodiment uses the benchmark data set TI46, and the speech signal is preprocessed into a pulse sequence through the Lyon ear model and the BSA algorithm. After the input layer neurons receive the speech pulse sequence, they transmit the signal to the brain-like model in the reserve layer through the synaptic model in the form of synaptic current. At this time, the neuron model will receive two types of input currents, namely the input current containing the speech or image signal transmitted from the input layer neuron model and the synaptic current transmitted from other neurons connected to it in the brain-like model. When the pulse sequence is input, the mathematical description of the input current of the neurons in the reserve layer brain-like model is as follows:

[0041]

[0042] Where, I si (t) represents the sum of the input currents from the input layer containing speech signals transmitted to neuron i; I gi (t) represents the sum of synaptic currents from other neurons in the brain-like model transmitted to neuron i; w si represents the sum of the input synaptic weights connected to the i-th neuron; represents the impulse response of the i-th neuron, that is, the change in membrane potential compared to the recovery potential when the neuron discharges.

[0043] After the input current transmitted from the input layer is received by the neuron model in the brain-like model, the synaptic plasticity model in the network performs adaptive adjustment, thereby changing the neuron discharge of the brain-like model. Therefore, different signals are input into the brain-like model to form their own neuron discharge patterns, and the discharge patterns can be used as identification features.

[0044] The characteristic signals extracted by the brain-inspired model are transmitted to the output layer neuron model through the synaptic plasticity mechanism. The synaptic connections between the reserve layer and the output layer are weighted using the remote supervised learning algorithm (ReSuMe). The ReSuMe algorithm has significant advantages in time series processing and is also biologically interpretable. This method is based on the STDP rule and adaptively optimizes synaptic weights by adjusting the discharge activity of neurons, effectively capturing the temporal dependency between input and output. The ReSuMe algorithm regulates the temporal changes of synaptic weights as follows:

[0045]

[0046] Where, S(t in ) represents the input pulse sequence; S(t a ) represents the actual output pulse sequence; S(t d ) represents the expected output pulse sequence; e Hrepresents the weight correction of the anti-Hebbian term; w(Δt) represents the correction function of the STDP rule; S(t in ) represents the pulse sequence of the neural model in the brain-like model; S(t a ) represents the actual output pulse sequence of the output layer neuron model after each iteration cycle; S(t d ) represents the expected output pulse sequence for each signal, which is determined according to the different discharge patterns of the brain-like model induced by different signals; e H is set to 0.25.

[0047] After training, the synaptic plasticity model shows that at the desired spike firing time, the average synaptic weight between the reserve layer neurons and the output layer neurons exhibits a periodic alternation of positive and negative spikes. When the synaptic weight reaches a positive peak, the spike firing activity of the output layer neurons is significantly enhanced. This regulatory mechanism causes the actual firing sequence of the output layer neurons to gradually approach the target spike sequence, thereby influencing the firing of the output layer neurons and making decisions based on the output neuron firing results.

[0048] Step S103: Apply electromagnetic intervention to different brain regions of the brain-like model, determine the optimal electromagnetic intervention parameters by analyzing the speech recognition accuracy of the brain-like model before and after the electromagnetic intervention, and perform brain-like model speech recognition based on the optimal electromagnetic intervention parameters.

[0049] Specifically, in this embodiment, electromagnetic interventions of different intensities and frequencies were applied to the prefrontal lobe, hippocampus and whole brain areas respectively, and the regulatory effect of electromagnetic intervention on the external presentation performance of the brain-like model was analyzed with recognition accuracy as the evaluation index.

[0050] AC magnetic fields affect the discharge activity of neurons through the induced electric field. When neurons are exposed to an AC magnetic field, according to Faraday's law of electromagnetic induction, the alternating magnetic field induces an alternating electric field around the neurons, which is mathematically described as follows:

[0051]

[0052] Where E(t) represents the induced electric field induced by the alternating magnetic field; r represents the radius of the alternating magnetic field; and B(t) represents the magnetic flux density. According to Coulomb's law, the induced electric field can cause charged ions to migrate, and charge to accumulate in certain parts of the cell membrane. This increased charge accumulation can lead to depolarization of the neuronal membrane potential. The relationship between the induced electric field E(t) and the resulting depolarization-induced voltage ΔV(t) is shown below:

[0053]

[0054] Where, t M represents the Maxwell-Wagner time constant; ld The polarization length describes the distance or length that the potential difference on a neuron can maintain when it propagates along the axon or dendrite. When the applied AC magnetic field is B(t) = Asin(2πft), the induced cell membrane depolarization voltage ΔV(t) is as follows:

[0055]

[0056] Where A and f represent the amplitude and frequency of the applied AC magnetic field; θ represents the angle between the direction of the external magnetic field and the normal vector of the neuron. M The magnitude is generally 10-10 and the AC magnetic field is in the low frequency range, so 2πft<<1, and formula (5) can be simplified as follows:

[0057] ΔV(t)≈πrfAl d cos(2πft) (6)

[0058] Therefore, the change of neuronal membrane potential v(t) under the action of AC magnetic field can be expressed as follows:

[0059] v(t)←v(t)+πrfAl d cos(2πft) (7)

[0060] Substituting formula (7) into the neuron model, we can obtain the discharge sequence of the prefrontal lobe neuron model and hippocampal neuron model under the action of the AC magnetic field B(t) = Asin(2πft). Taking the electromagnetic control of 10Hz / 0.5T as an example, the applied magnetic field waveform is as follows: Figure 3 shown.

[0061] Based on the above content, the specific application process of electromagnetic regulation in this embodiment is as follows: electromagnetic interventions of different intensities and frequencies are applied to all neuron models in the prefrontal lobe brain region, all neuron models in the hippocampus brain region, and simultaneously applied to all neuron models in the prefrontal lobe and hippocampus brain regions, that is, the cell membrane depolarization induced voltage ΔV(t) caused by electromagnetic regulation is applied to the membrane potential of the prefrontal lobe and hippocampus neuron models according to formula (7).

[0062] The electromagnetic control parameters ranged from [0, 1] T, with a step size of 0.1 T; the applied magnetic field intensity ranged from 1 Hz, 10 Hz, and 20 Hz; and the magnetic field stimulation duration was 1000 ms. The results showed that a certain range of electromagnetic control parameters could improve the performance of the brain-inspired model, and that there was an optimal electromagnetic control parameter that maximized the performance of the brain-inspired model.

[0063] Electromagnetic intervention was applied to the brain-like model. The speech recognition accuracy of different brain regions (prefrontal lobe, hippocampus, prefrontal lobe and hippocampus) of the brain-like model with different frequencies and amplitudes of electromagnetic intervention was as follows: Figure 4 As shown, Figure 4 The middle black dashed line represents the speech recognition accuracy of the brain-inspired model without electrical intervention. When subjected to electromagnetic intervention, the overall change in recognition accuracy shows a trend of first increasing and then decreasing, indicating that electromagnetic parameter intervention within an appropriate range can improve the speech recognition accuracy of the brain-inspired model. The electromagnetic control parameters that resulted in the greatest improvement in the recognition accuracy of the brain-inspired model were: 10 Hz / 0.2 T, when electromagnetic intervention simultaneously stimulated the prefrontal cortex and hippocampus.

[0064] Based on the above analysis, the speech recognition performance of the brain-inspired model can be improved under the stimulation of a certain range of electromagnetic intervention parameters. The electromagnetic intervention parameters that achieve the highest recognition accuracy are determined to be the optimal electromagnetic intervention parameters. This embodiment can further enhance the biological interpretability and neural information processing capabilities of brain-inspired models and promote the development of brain-inspired intelligence in pattern recognition tasks.

[0065] The method described in this embodiment can effectively improve the speech recognition performance of the brain-like model by applying a determined optimal electromagnetic intervention, further enhance the biological interpretability and neural information processing capabilities of the brain-like model, and promote the development of brain-like intelligence in pattern recognition tasks.

[0066] It should be noted that the above description is limited to some embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in an order different from that described in the above embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0067] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, an embodiment of the present application further provides a speech recognition device based on a brain-like model.

[0068] like Figure 5 As shown, the speech recognition device based on the brain-like model includes:

[0069] The model construction module 11 is configured to obtain a whole-brain network topology structure based on the brain function network generated by human brain imaging data, and to construct a multi-brain region spiking neural network as a brain-like model using multiple types of neuron models as nodes and synaptic plasticity models as edges;

[0070] a speech recognition framework construction module 12 configured to construct a speech recognition framework of a brain-inspired model, wherein the speech recognition framework includes input layer neurons, a reserve layer model, and output layer neurons, takes a speech pulse sequence as input, and trains synaptic weights between the reserve layer model and the output layer neurons using a remote supervised learning algorithm;

[0071] The electromagnetic intervention module 13 is configured to apply electromagnetic intervention to different brain regions of the brain-like model, determine the optimal electromagnetic intervention parameters by analyzing the speech recognition accuracy of the brain-like model before and after the electromagnetic intervention, and perform brain-like model speech recognition based on the optimal electromagnetic intervention parameters.

[0072] For the convenience of description, the above devices are described as being divided into various modules according to their functions. Of course, when implementing the embodiments of the present application, the functions of each module can be implemented in the same or multiple software and / or hardware.

[0073] The apparatus of the above embodiment is used to implement the corresponding method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be described in detail here.

[0074] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, an embodiment of the present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and runnable on the processor, wherein when the processor executes the program, the method described in any of the above embodiments is implemented.

[0075] Figure 6 10 is a schematic diagram showing a more specific hardware structure of an electronic device provided in this embodiment. The device may include: a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040 are communicatively connected to each other within the device via the bus 1050.

[0076] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0077] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage devices, dynamic storage devices, etc. The memory 1020 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.

[0078] The input / output interface 1030 is used to connect an input / output module to implement information input and output. The input / output module can be configured as a component in the device (not shown in the figure) or can be externally connected to the device to provide corresponding functions. Input devices may include a keyboard, mouse, touch screen, microphone, various sensors, etc., and output devices may include a display, speaker, vibrator, indicator light, etc.

[0079] The communication interface 1040 is used to connect to a communication module (not shown) to enable communication between the device and other devices. The communication module can communicate via a wired method (such as USB, network cable, etc.) or a wireless method (such as mobile network, WiFi, Bluetooth, etc.).

[0080] The bus 1050 comprises a path for transmitting information between the various components of the device (eg, the processor 1010 , the memory 1020 , the input / output interface 1030 , and the communication interface 1040 ).

[0081] It should be noted that although the above device only shows the processor 1010, the memory 1020, the input / output interface 1030, the communication interface 1040, and the bus 1050, in a specific implementation, the device may also include other components necessary for normal operation. In addition, it will be understood by those skilled in the art that the above device may only include the components necessary to implement the embodiments of this specification, and does not necessarily include all the components shown in the figure.

[0082] The electronic device of the above embodiment is used to implement the corresponding method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be described in detail here.

[0083] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, the present application also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the method described in any of the above embodiments.

[0084] The computer-readable media of this embodiment include permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0085] The computer instructions stored in the storage medium of the above embodiment are used to enable the computer to execute the method described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0086] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present application (including the claims) is limited to these examples. Within the scope of the present application, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the embodiments of the present application as described above, which are not provided in detail for the sake of simplicity.

[0087] In addition, for simplicity of description and discussion, and in order not to make the embodiment of the application difficult to understand, the known power supply / ground connection with integrated circuit (IC) chip and other components may or may not be shown in the accompanying drawings provided. In addition, the device can be shown in the form of a block diagram to avoid making the embodiment of the application difficult to understand, and this also takes into account the following fact, that is, the details of the embodiment of these block diagram devices are highly dependent on the platform to be implemented in the embodiment of the application (that is, these details should be fully within the scope of understanding of those skilled in the art). When specific details (for example, circuit) are set forth to describe exemplary embodiments of the application, it will be apparent to those skilled in the art that the embodiment of the application can be implemented without these specific details or when these specific details are changed. Therefore, these descriptions should be considered to be illustrative rather than restrictive.

[0088] Although the present invention has been described in conjunction with specific embodiments thereof, many alternatives, modifications, and variations of these embodiments will be apparent to those skilled in the art based on the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may utilize the embodiments discussed.

[0089] The embodiments of the present application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present application should be included in the scope of protection of this application.

Claims

1. A speech recognition method based on a brain-like model, characterized in that: include: The brain functional network generated from human brain imaging data was used to obtain the whole-brain network topology. A multi-brain region spiking neural network was constructed as a brain-like model using multiple types of neuron models as nodes and synaptic plasticity models as edges. Constructing a speech recognition framework based on a brain-inspired model, wherein the speech recognition framework includes input layer neurons, a reservoir layer model, and output layer neurons, takes a speech pulse sequence as input, and trains the synaptic weights between the reservoir layer model and the output layer neurons using a remote supervised learning algorithm; Electromagnetic intervention is applied to different brain regions of the brain-like model, and optimal electromagnetic intervention parameters are determined by analyzing the speech recognition accuracy of the brain-like model before and after the electromagnetic intervention, and brain-like model speech recognition is performed according to the optimal electromagnetic intervention parameters.

2. The method according to claim 1, characterized in that The brain function network generated according to the human brain imaging data obtains a whole-brain network topology structure, including: Locate the brain regions according to the brain region numbers in the brain atlas, and divide the brain functional network into multiple functional brain regions to obtain a multi-brain region structure; The functional connectivity strength of the time series of physiological and functional metabolic measurement signals of network nodes is defined as the edge of the network. According to the correlation coefficient between the determined brain region nodes and based on the network topology threshold, the functional connectivity matrix between the nodes of each brain region is obtained to obtain the whole-brain network topology structure.

3. The method according to claim 1, wherein: The dedicated neuron model is configured to the corresponding brain region nodes, and the remaining brain region nodes are configured with the general neuron model.

4. The method according to claim 1, wherein: The synaptic plasticity model includes excitatory synapses and inhibitory synapses, and introduces random synaptic delays to characterize the diffusion of neurotransmitters in biological synapses.

5. The method according to claim 1, wherein: Preprocessing the speech signal into a speech pulse sequence; after receiving the speech pulse sequence, the input layer neurons transmit the signal to the storage layer model in the form of synaptic current through the synaptic model, forming respective neuron discharge patterns, and the discharge patterns serve as recognition features; The extracted feature signals are transmitted to the output layer neurons through the synaptic plasticity mechanism.

6. The method according to claim 1, wherein: The synaptic connection between the reservoir layer model and the output layer neurons is weighted by a remote supervised learning algorithm. The remote supervised learning algorithm regulates the change of synaptic weight over time as follows: Where, S(t in ) represents the input pulse sequence; S(t a ) represents the actual output pulse sequence; S(t d ) represents the expected output pulse sequence; e H represents the weight correction of the anti-Hebbian term; w(Δt) is the correction function of the STDP rule. S(t in ) represents the pulse sequence of the neural model in the brain-like model; S(t a ) represents the actual output pulse sequence of the output layer neuron model after each iteration cycle; S(t d ) represents the expected output pulse sequence for each signal.

7. The method according to claim 1, wherein: Electromagnetic interventions of different intensities and frequencies were applied to the prefrontal lobe, hippocampus, and whole brain regions, and the recognition accuracy was used as the evaluation index to analyze the regulatory effect of electromagnetic intervention on the external presentation performance of the brain-like model. The changes in neuronal membrane potential under the action of the AC magnetic field are expressed as: v(t)←v(t)+πrfAl d cos(2πft); Where r is the radius of the alternating magnetic field; A and f are the amplitude and frequency of the applied alternating magnetic field, respectively; l d represents polarization length; v(t) represents membrane potential; t represents time.

8. A speech recognition device based on a brain-like model, characterized in that: include: A model building module is configured to obtain a whole-brain network topology structure based on a brain functional network generated from human brain imaging data, and to construct a multi-brain region spiking neural network as a brain-like model using multiple types of neuron models as nodes and synaptic plasticity models as edges; a speech recognition framework construction module configured to construct a speech recognition framework of a brain-inspired model, wherein the speech recognition framework includes input layer neurons, a reservoir layer model, and output layer neurons, takes a speech pulse sequence as input, and trains synaptic weights between the reservoir layer model and the output layer neurons using a remote supervised learning algorithm; The electromagnetic intervention module is configured to apply electromagnetic intervention to different brain regions of the brain-like model, determine the optimal electromagnetic intervention parameters by analyzing the speech recognition accuracy of the brain-like model before and after the electromagnetic intervention, and perform brain-like model speech recognition based on the optimal electromagnetic intervention parameters.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 7 when executing the program.

10. A non-transitory computer-readable storage medium, characterized in that in, The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to cause a computer to execute the method according to any one of claims 1 to 7.

Images