Autistic child emotion recognition and adjustment method and system based on wearable device

Abstract

The invention belongs to the field of autistic child emotion regulation, and provides a wearable device-based autistic child emotion recognition and regulation method and system, and the method comprises the following steps: carrying out the data processing of the physiological data and facial expression of an autistic child, inputting the processed data into a hybrid multilayer neural network, and obtaining an emotion recognition result; if the emotion is recognized as the negative emotion, the inflatable vest is inflated to relieve the negative emotion of the child patient. According to the method, the emotion of the self-closing child is recognized by using the multi-modal data, the real emotion characteristics of the child patient can be fully mined, and the emotion recognition accuracy is improved. Meanwhile, by means of the sensitivity of physiological signals to emotional fluctuation, the effects of emotion early warning and advanced intervention are achieved. In addition, various wearable devices are systematically integrated, and the emotion of the autistic children can be continuously, intelligently and effectively adjusted.

Description

Technical Field

[0001] The present invention belongs to the field of emotion regulation for children with autism, and specifically relates to fields such as wearable smart devices, deep learning algorithm application, physiological signal processing, deep pressure therapy (DPT), and especially a method and system for emotion recognition and regulation for children with autism based on wearable devices. Technical Background

[0002] At present, when autistic children have negative emotions, they can often only detect them through observation of their body, facial expressions and other external manifestations. This delay in emotional perception greatly hinders timely emotional intervention and seriously affects the patient's recovery process. Current research shows that body physiological signals (such as skin electrical response, heart rate, etc.) can be identified in the early stages of emotional fluctuations, which is more forward-looking in predicting emotions.

[0003] At the same time, traditional intervention methods (such as support and companionship) usually intervene after negative emotions appear, and cannot achieve early intervention. In addition, although deep pressure inflatable vests based on deep pressure therapy (DPT) have been shown to relieve anxiety, anger and other negative emotions in children with autism, manually inflated pressure vests lack intelligence and ease of use.

[0004] Therefore, based on the above problems, there is an urgent need for a more intelligent and accurate method and system for emotion recognition and regulation in children with autism, which can provide more timely and effective emotion monitoring and regulation based on the special needs of children with autism. Summary of the invention

[0005] In view of the problems and shortcomings of the prior art, the present invention provides a method and system for emotion recognition and regulation of autistic children based on wearable devices.

[0006] The present invention solves the above technical problems through the following technical solutions:

[0007] The present invention provides a method for emotion recognition and regulation of autistic children based on a wearable device, which comprises the following steps:

[0008] Step S1, obtaining in real time the physiological signals of the current user within a certain time period, including skin electrical signals, skin temperature signals, heart rate signals, blood volume pulse signals, and simultaneously obtaining in real time the facial expression image of the current user within the corresponding time period;

[0009] Step S2, performing a noise reduction filter preprocessing operation on the acquired physiological signal, and performing a preprocessing operation on the acquired facial expression image;

[0010] Step S3, inputting the pre-processed physiological signal data and facial expression image data into a hybrid multi-layer neural network for feature extraction, emotion recognition and classification, and outputting the emotion classification judgment result;

[0011] Step S4, determining whether the emotion classification result in step S3 is a negative emotion, if the emotion classification is a negative emotion, proceeding to step S5, if the emotion classification is not a negative emotion, proceeding to step S6;

[0012] Step S5, determining whether the deep pressure inflatable vest is in an inflated state, if the deep pressure inflatable vest is in an inflated state, proceeding to step S8, if the deep pressure inflatable vest is not in an inflated state, proceeding to step S7;

[0013] Step S6, determining whether the deep pressure inflatable vest is in an inflated state, if the deep pressure inflatable vest is in an inflated state, proceeding to step S9, if the deep pressure inflatable vest is not in an inflated state, returning to step S1 to repeat the entire method process, and continuing to obtain and classify the emotional state;

[0014] Step S7, the deep pressure inflatable vest is inflated, and the method returns to step S1 to repeat the entire process, and continues to obtain and classify the emotional state until the emotional state reaches a normal index, enabling the deep pressure inflatable vest to be deflated;

[0015] Step S8, continuing to maintain the inflated state of the deep pressure inflatable vest, while returning to step S1 to repeat the entire method process, continuing to obtain and classify the emotional state, until the emotion reaches a normal index, enabling the deep pressure inflatable vest to be deflated;

[0016] In step S9, the deep pressure inflatable vest is deflated, and the process returns to step S1 to repeat the entire method process to continue acquiring and classifying the emotional state.

[0017] The present invention also provides a system for emotion recognition and regulation of autistic children based on wearable devices, which includes an acquisition module, a data analysis module, a wireless communication module, and an adjustment module. The data analysis module is connected to the wireless communication module, the acquisition module is connected to the data analysis module and the wireless communication module respectively, the adjustment module is connected to the wireless communication module, and the wireless communication module is connected to a terminal;

[0018] An acquisition module is used to obtain data reflecting the user's emotional state in real time within a certain period of time;

[0019] The data analysis module is used to pre-process and analyze the data collected by the acquisition module to determine and output the emotion category;

[0020] A wireless communication module, including one or more of a network controller, a Bluetooth transceiver, a WiFi transceiver, and a mobile network transceiver, for transmitting the collected data of the acquisition module to the data analysis module, and transmitting the action instructions generated by the data analysis results to the adjustment module;

[0021] The regulation module is used to provide timely emotion regulation measures when the data analysis module analyzes that the autistic children have negative emotions. The regulation module is triggered by the results of the data analysis module. When the data analysis module identifies a potential negative emotional state, it will activate or keep the regulation module running. When the data analysis module identifies a non-potential negative emotional state, it will terminate the regulation module.

[0022] The terminal is used to receive data analysis results and adjustment module action status from the wearable device, and provide an intuitive user interface so that the user can understand the received information.

[0023] Further, in order to ensure the data collection effect, the acquisition module includes a facial collection module and a physiological signal collection module, the facial collection module is connected to the data analysis module, and the physiological signal collection module is connected to the data analysis module;

[0024] A facial acquisition module is used to collect the user's facial expression image data;

[0025] The physiological signal acquisition module is used to collect the physiological signal data of the user.

[0026] Furthermore, in order to ensure the effect of facial data expression collection, a high-definition camera is used in the facial collection module, and the high-definition camera is connected to the data analysis module to obtain the user's facial expression image in real time.

[0027] Furthermore, in order to ensure the physiological signal data collection effect, a health monitoring smart bracelet is used in the physiological signal collection module, and the health monitoring smart bracelet is connected to a data analysis module;

[0028] The health monitoring smart bracelet is used to monitor and obtain the user's four physiological signals of skin electricity, skin temperature, heart rate and blood volume pulse in real time.

[0029] Furthermore, the data analysis module includes a data processing unit, a hybrid multi-layer neural network, and a command and dispatch unit;

[0030] The data processing unit is used to obtain the time-frequency domain analysis features corresponding to the four physiological signals of skin electricity, skin temperature, heart rate, and blood volume pulse, and to convert the resolution and depth of the collected facial expression images;

[0031] A hybrid multi-layer neural network is used to extract features and identify emotions from pre-processed data to obtain emotion classification results, which can ensure the processing effect of the data;

[0032] The command and dispatch unit is used to provide command, dispatch and coordination data or instructions to networked wearable devices and terminals.

[0033] Furthermore, in order to ensure the data transmission effect, the wireless communication module uses a wireless Bluetooth transceiver, a WiFi transceiver and a mobile network transceiver for data communication and transmission.

[0034] Furthermore, in order to ensure the effect of negative emotion regulation, a deep pressure inflatable vest is used in the regulation module;

[0035] Deep pressure inflatable vest, a wearable vest based on deep pressure therapy DPT to relieve negative emotions such as anxiety. In the system, it is used to receive inflation and deflation instructions from the data analysis module to execute or terminate the user's emotion regulation process.

[0036] Furthermore, the data processing unit is used to perform low-pass filtering and denoising on the four physiological signals of skin electricity, skin temperature, heart rate and blood volume pulse input by the acquisition module, and to process and acquire relevant time-frequency domain analysis features, while converting the resolution and depth of the facial expression image.

[0037] Furthermore, the hybrid multi-layer neural network can ensure the effect of emotion recognition. It uses LSTM to extract the time-frequency analysis features of the four physiological signals after data preprocessing, and uses VGG-16 to extract the features of the preprocessed facial expression images. Then, the convolution operation is used to transform the two-part feature vectors to obtain the fused features, and then the attention mechanism is used to analyze the sequence and capture the connection between these features. Finally, the emotion classification judgment result is output through the fully connected layer.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] By setting up the acquisition module and the data analysis module, and the mutual cooperation between the two modules and the regulation module, the wireless communication module, and the terminal, real-time emotional monitoring and regulation of autistic children can be achieved. The acquisition module can accurately capture the patient's facial behavior and physiological state, and the data analysis module classifies the collected data into emotions to identify the possibility that the patient is in a negative emotional state. Once the system identifies a negative emotional state, the regulation module cooperates with the wireless communication module and the terminal to take emotional regulation measures in advance and in time through the deep pressure inflatable vest. This mutually coordinated monitoring mechanism not only improves the accuracy and efficiency of monitoring, but also provides patients with more timely and effective emotional intervention support, which helps to better manage emotional states and prevent emotional breakdowns. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 This is a flow chart of a method for emotion recognition and regulation of autistic children based on wearable devices according to an embodiment of the present invention.

[0042] Figure 2 4 is a system structure block diagram of emotion recognition and regulation for autistic children based on wearable devices according to an embodiment of the present invention.

[0043] Figure 3 is a structural diagram of a hybrid multi-layer neural network according to an embodiment of the present invention.

[0044] Figure 4 2 is a schematic structural diagram of a deep pressure inflatable vest according to an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0046] like Figure 1 As shown, this embodiment provides a method for emotion recognition and regulation of autistic children based on a wearable device, which includes the following steps:

[0047] Step S1, real-time acquisition of the physiological signals of the current user within a certain period of time, including skin electrical signals, skin temperature signals, heart rate signals, blood volume pulse signals, and real-time acquisition of the facial expression image of the current user within a certain period of time.

[0048] The "certain time period" cannot be too short or too long. If the time is too short, it will have a potential impact on the accuracy of emotion recognition. If the time is too long, negative emotions may not be regulated and intervened in time. It is generally limited to t seconds, 30<t<60, and in this embodiment, it is set to 30 seconds.

[0049] Step S2, performing a noise reduction filter preprocessing operation on the acquired physiological signal, and simultaneously performing a preprocessing operation on the acquired facial expression image.

[0050] In this embodiment, the four physiological signals of skin electricity, skin temperature, heart rate and blood volume pulse obtained are subjected to low-pass filtering and denoising. The low-pass filtering range of the skin electricity signal is [0, 0.08] Hz, the low-pass filtering range of the skin temperature is [0, 0.2] Hz, the low-pass filtering range of the heart rate is [0, 0.4] Hz, and the low-pass filtering range of the blood volume pulse is [0, 15] Hz. The time domain analysis features of the skin electricity signal include the mean, standard deviation, median, maximum and ratio of the skin electricity signal, the mean, standard deviation, median, maximum and ratio of the skin electricity signal after the first-order difference, the mean, standard deviation, median, maximum and ratio of the skin electricity signal after the second-order difference, the frequency domain analysis features of the skin electricity signal include the spectral power of the frequency band [0, 0.08] Hz, the time domain analysis features of the skin temperature signal The features include the mean, standard deviation, median, maximum and ratio of the electrodermal signal; the frequency domain analysis features of the skin temperature signal include the spectral power in the frequency band [0, 0.2] Hz; the time domain analysis features of the heart rate signal include the heart rate mean, heart rate standard deviation, heart rate variability HRV mean and heart rate variability HRV standard deviation; the frequency domain analysis features of the heart rate signal include the power spectral density in the frequency band [0, 0.4] Hz; the time domain analysis features of the blood volume pulse signal include the mean, standard deviation, median, maximum and ratio of the blood volume pulse signal; the frequency domain analysis features of the blood volume pulse signal include the power spectral density in the frequency band [0, 0.15] Hz; at the same time, the acquired facial expression image is converted into a [224×224] resolution, and the image depth is converted from 24 bits to 8 bits.

[0051] Step S3, inputting the preprocessed physiological signal data and facial expression data together into a hybrid multi-layer neural network for feature extraction, emotion recognition and classification, and outputting the emotion classification judgment result.

[0052] The structure of the hybrid multi-layer neural network used in this embodiment is as follows Figure 3As shown, LSTM (Long Short-Term Memory) neural network is used to process physiological signal data that changes over time. The time-frequency analysis features of four physiological signals are extracted through LSTM neural network. At the same time, VGG-16 convolutional neural network is used to process the preprocessed facial expression images layer by layer to extract high-dimensional features related to emotions. Then, MCB (Multimodal Compact Bilinear Pooling) is used to fuse the LSTM output from physiological signals and the VGG-16 output of facial expression images to obtain a joint feature vector representing two different modalities. The fused features are then given different feature weights through the attention mechanism module to enhance the focus on key features. Finally, the output features are again passed through a fully connected layer of an LSTM neural network and a Softmax activation function to map the features to the probability distribution of different emotion categories. Finally, the emotion category with the highest probability is used as the recognition and classification result. The emotion classification judgment results are set as positive emotions (happy), neutral emotions (calm) and negative emotions (anger, anxiety, sadness).

[0053] Step S4, determining whether the emotion classification result in step S3 is a negative emotion, if "yes", proceed to step S5, if "no", proceed to step S6.

[0054] Step S5, under the premise that the emotion classification judgment result in step S4 is "yes", determine whether the deep pressure inflatable vest is in an inflated state, if "yes" proceed to step S8, if "no" proceed to step S7.

[0055] Step S6, if the result of emotion classification judgment in step S4 is "no", determine whether the deep pressure inflatable vest is inflated, if "yes", proceed to step S9, if "no", return to step S1 and repeat the whole method process to continue to obtain and classify the emotion state;

[0056] Step S7, under the premise that the emotion classification judgment result in step S4 is "yes" and the inflation state of the deep pressure inflatable vest in step S5 is "no", the deep pressure inflatable vest is inflated, and at the same time returns to step S1 to repeat the entire method process, and continues to obtain and classify the emotional state until the emotion reaches the normal indicator, enabling the deep pressure inflatable vest to be deflated.

[0057] If the emotion classification result is one of the negative emotions (anger, anxiety, sadness), and the deep pressure inflatable vest is in a non-inflated state, the deep pressure inflatable vest will produce an inflation action.

[0058] The deep pressure inflatable vest structure used in this embodiment is as follows Figure 4 As shown, a pump release valve device is used to perform the inflation and deflation actions.

[0059] Step S8, under the premise that the emotion classification judgment result in step S4 is "yes" and the inflation state of the deep pressure inflatable vest in step S5 is "yes", continue to maintain the inflation state of the deep pressure inflatable vest, and return to step S1 to repeat the entire method process, continue to obtain and classify the emotion state, and enable the deep pressure inflatable vest to be deflated after the emotion reaches the normal index;

[0060] This embodiment takes into account that when the user's emotion classification judgment results are both negative emotions (anger, anxiety, sadness) twice in a continuous time period, that is, the deep pressure inflatable vest has been triggered to be inflated for the previous time, the deep pressure inflatable vest will still remain inflated for the next time.

[0061] Step S9, under the premise that the emotion classification judgment result in step S4 is "no" and the inflation state of the deep pressure inflatable vest in step S6 is "yes", the deep pressure inflatable vest is deflated and returns to step S1 to repeat the entire method process to continue to obtain and classify the emotional state.

[0062] This embodiment takes into account that when the user's emotion classification results are negative emotions (anger, anxiety, sadness) and positive emotions (happiness) or neutral emotions (calm) twice in a continuous time period, that is, the deep pressure inflatable vest has been triggered to be inflated, the emotion regulation state is released in the next time, and the deep pressure inflatable vest performs a deflation action.

[0063] like Figure 2 As shown, this embodiment also provides a system for emotion recognition and regulation of autistic children based on wearable devices, which includes an acquisition module, a data analysis module, a wireless communication module, and an adjustment module. The data analysis module is connected to the wireless communication module, the acquisition module is connected to the data analysis module and the wireless communication module respectively, the adjustment module is connected to the wireless communication module, and the wireless communication module is connected to the terminal;

[0064] The acquisition module is used to obtain data reflecting the user's emotional state in real time within a certain period of time.

[0065] The data analysis module is used to analyze the various data collected by the acquisition module and determine the emotion category.

[0066] The wireless communication module includes one or more of a network controller, a Bluetooth transceiver, a WiFi transceiver and a mobile network transceiver, and is used to transmit the collected data of the acquisition module to the data analysis module, and transmit the action instructions generated by the data analysis results to the adjustment module.

[0067] Among them, one or more of the network controller, Bluetooth transceiver, WiFi transceiver and mobile network transceiver depends on the device used. The high-definition camera and health monitoring smart bracelet of this embodiment use the 5G networking module and Bluetooth 5.0 module that come with the device, and the deep pressure inflatable vest uses the Bluetooth 5.0 module for command action driving;

[0068] The WiFi transceiver realizes medium and short distance wireless networking. This embodiment adopts a 1000M WiFi wireless transceiver.

[0069] The regulation module is used to provide timely emotion regulation measures when the data analysis module finds that autistic children have negative emotions.

[0070] The terminal is used to receive data analysis results and adjustment module action status from the wearable device, and provide an intuitive user interface so that the user can understand the received information.

[0071] Specifically, the specific process of each module in the system of this embodiment and the module realizing its function can be found in the relevant description in the method embodiment, which will not be repeated here.

[0072] This embodiment also provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute the above-mentioned method for identifying and regulating emotions of autistic children based on wearable devices.

[0073] Scenario description:

[0074] For example, a child with autism has recently been emotionally unstable and often suffers from negative emotions such as irritability and restlessness. The doctor recommends using our wearable device to monitor his emotional state so that he can regulate his negative emotions in a timely manner.

[0075] Feature collection:

[0076] Facial expression collection: Use the high-definition camera module of the wearable device to capture the user's facial micro-expression changes. For example, when the user is angry, anxious or sad, his face may show slight expression changes. These data are analyzed by the VGG-16 neural network to assist in judging his emotional state.

[0077] Physiological signal data collection: Use a health detection smart bracelet to continuously monitor the user's physiological indicators to collect real-time data on skin electrical signals, skin temperature signals, heart rate signals, blood volume and pulse signals. These data can reflect the user's physiological and emotional responses in advance.

[0078] Emotion recognition and regulation:

[0079] All collected data will be transmitted to a data analysis module, which includes a data processing unit, a command and dispatch unit, etc. These modules work together to conduct a comprehensive analysis of the data to determine whether the user is in a potential negative emotional state.

[0080] If the system determines that the user is in a negative emotional state, the adjustment module will be triggered. This module uses a deep pressure inflatable vest to apply pressure to make the user feel like being hugged, namely deep pressure therapy DPT, thereby calming their negative emotions.

[0081] Although the specific embodiments of the present invention are described above, those skilled in the art should understand that these are only examples, and the protection scope of the present invention is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but these changes and modifications all fall within the protection scope of the present invention.

Claims

1. A method for emotion recognition and regulation of autistic children based on wearable devices, characterized in that: It includes the following steps: Step S1, obtaining in real time the physiological signals of the current user within a certain time period, including skin electrical signals, skin temperature signals, heart rate signals, blood volume pulse signals, and simultaneously obtaining in real time the facial expression image of the current user within the corresponding time period; Step S2, performing a noise reduction filter preprocessing operation on the acquired physiological signal, and performing a preprocessing operation on the acquired facial expression image; Step S3, inputting the pre-processed physiological signal data and facial expression image data into a hybrid multi-layer neural network for feature extraction, emotion recognition and classification, and outputting the emotion classification judgment result; Step S4, determining whether the emotion classification result in step S3 is a negative emotion, if the emotion classification is a negative emotion, proceeding to step S5, if the emotion classification is not a negative emotion, proceeding to step S6; Step S5, determining whether the deep pressure inflatable vest is in an inflated state, if the deep pressure inflatable vest is in an inflated state, proceeding to step S8, if the deep pressure inflatable vest is not in an inflated state, proceeding to step S7; Step S6, determining whether the deep pressure inflatable vest is in an inflated state, if the deep pressure inflatable vest is in an inflated state, proceeding to step S9, if the deep pressure inflatable vest is not in an inflated state, returning to step S1 to repeat the entire method process, and continuing to obtain and classify the emotional state; Step S7, the deep pressure inflatable vest is inflated, and the method returns to step S1 to repeat the entire process, and continues to obtain and classify the emotional state until the emotional state reaches a normal index, enabling the deep pressure inflatable vest to be deflated; Step S8, continuing to maintain the inflated state of the deep pressure inflatable vest, while returning to step S1 to repeat the entire method process, continuing to obtain and classify the emotional state, until the emotion reaches a normal index, enabling the deep pressure inflatable vest to be deflated; In step S9, the deep pressure inflatable vest is deflated, and the process returns to step S1 to repeat the entire method process to continue acquiring and classifying the emotional state.

2. According to claim 1, a method for emotion recognition and regulation of autistic children based on wearable devices is characterized in that: In step S2, four physiological signals, namely, skin electricity, skin temperature, heart rate, and blood volume pulse, are subjected to low-pass filtering and denoising; wherein: The time domain analysis features of the skin electrical signal include the mean, standard deviation, median, maximum and ratio of the skin electrical signal, the mean, standard deviation, median, maximum and ratio of the skin electrical signal after the first-order difference, the mean, standard deviation, median, maximum and ratio of the skin electrical signal after the second-order difference, and the frequency domain analysis features of the skin electrical signal include spectral power; The time domain analysis characteristics of the skin temperature signal include the mean, standard deviation, median, maximum and ratio of the skin electrical signal, and the frequency domain analysis characteristics of the skin temperature signal include spectral power; The time domain analysis features of the heart rate signal include the heart rate mean, heart rate standard deviation, heart rate variability HRV mean, heart rate variability HRV standard deviation, and the frequency domain analysis features of the heart rate signal include power spectrum density; The time domain analysis characteristics of the blood volume pulse signal include the mean, standard deviation, median, maximum and ratio of the blood volume pulse signal, and the frequency domain analysis characteristics of the blood volume pulse signal include power spectrum density.

3. A wearable device-based system for emotion recognition and regulation of autistic children, characterized by: It includes an acquisition module, a data analysis module, a wireless communication module, and an adjustment module. The data analysis module is connected to the wireless communication module, the acquisition module is connected to the data analysis module and the wireless communication module respectively, the adjustment module is connected to the wireless communication module, and the wireless communication module is connected to the terminal; An acquisition module is used to obtain data reflecting the user's emotional state in real time within a certain period of time; The data analysis module is used to pre-process and analyze the data collected by the acquisition module to determine and output the emotion category; A wireless communication module, including one or more of a network controller, a Bluetooth transceiver, a WiFi transceiver, and a mobile network transceiver, for transmitting the collected data of the acquisition module to the data analysis module, and transmitting the action instructions generated by the data analysis results to the adjustment module; The regulation module is used to provide timely emotion regulation measures when the data analysis module finds that the autistic child has negative emotions; The terminal is used to receive data analysis results and adjustment module action status from the wearable device, and provide an intuitive user interface so that the user can understand the received information.

4. A system for emotion recognition and regulation of autistic children based on wearable devices according to claim 3, characterized in that: The acquisition module includes a facial acquisition module and a physiological signal acquisition module respectively connected to the data analysis module; A facial acquisition module is used to collect the user's facial expression image data; The physiological signal acquisition module is used to collect the physiological signal data of the user.

5. A system for emotion recognition and regulation of autistic children based on wearable devices according to claim 4, characterized in that: The facial acquisition module includes a high-definition camera, which is connected to the data analysis module; High-definition camera, used to obtain real-time images of the user's facial expressions.

6. A system for emotion recognition and regulation of autistic children based on wearable devices according to claim 4, characterized in that: The physiological signal acquisition module includes a health monitoring smart bracelet, which is connected to the data analysis module; The health monitoring smart bracelet is used to monitor and obtain the user's four physiological signals of skin electricity, skin temperature, heart rate and blood volume pulse in real time.

7. A system for emotion recognition and regulation of autistic children based on wearable devices according to claim 3, characterized in that: The data analysis module includes a data processing unit, a hybrid multi-layer neural network, and a command and dispatch unit; The data processing unit is used to obtain the time-frequency domain analysis features corresponding to the four physiological signals of skin electricity, skin temperature, heart rate, and blood volume pulse, and to convert the resolution and depth of the collected facial expression images; A hybrid multi-layer neural network is used to perform feature extraction and emotion recognition operations on the preprocessed data to obtain emotion classification and judgment results; The command and dispatch unit is used to provide command, dispatch and coordination data or instructions to networked wearable devices and terminals.

8. A system for emotion recognition and regulation of autistic children based on wearable devices according to claim 3, characterized in that: The wireless communication module includes a wireless Bluetooth transceiver, a WiFi transceiver and a mobile network transceiver, which are used for data communication and transmission.

9. A system for emotion recognition and regulation of autistic children based on wearable devices according to claim 3, characterized in that: The regulation module includes a deep pressure inflatable vest; the deep pressure inflatable vest is a wearable vest based on deep pressure therapy DPT to relieve negative emotions such as anxiety. In the system, it can receive inflation and deflation instructions from the data analysis module to execute or terminate the user's emotion regulation process.

10. A wearable device-based autistic children's emotion recognition and regulation system according to any one of claims 7 to 9, characterized in that: A hybrid multi-layer neural network uses LSTM to extract the time-frequency analysis features of the four physiological signals after data preprocessing, and VGG-16 to extract the features of the preprocessed facial expression images. Then, the convolution operation is used to transform the two-part feature vectors to obtain the fused features. The attention mechanism is used to analyze the sequence and capture the connection between these features. Finally, the emotion classification judgment results are output through the fully connected layer.

Images