An Emotion Recognition Method and System Based on Physiological Parameter Extraction

Abstract

This invention discloses an emotion recognition method and system based on physiological parameter extraction, relating to the field of emotion recognition, and solving the problem of low accuracy in emotion recognition. The method includes setting emotion induction methods corresponding to different emotion categories, collecting multiple sets of electrocardiogram (ECG) signals from the test individual under different emotion categories; extracting features from the multiple sets of ECG signals from the test individual in both the time and frequency domains to obtain the time-domain and frequency-domain features corresponding to the ECG signals; using the time-domain and frequency-domain features corresponding to the multiple sets of ECG signals from the test individual as a training set to construct an emotion classification method; collecting the user's ECG signals within a recent time period and substituting them into the emotion classification method; generating and outputting the user's corresponding recognized emotion through the emotion classification method. This invention achieves accurate identification of user emotions by extracting features from the user's ECG signals and constructing a corresponding emotion classification method.

Description

Technical Field

[0001] This invention belongs to the field of emotion recognition technology, specifically an emotion recognition method and system based on physiological parameter extraction. Background Technology

[0002] Physiological parameters are quantitative indicators that reflect the physiological state and functional activities of the human body. They are obtained by measuring various biological signals of the human body in a resting or active state. These parameters can objectively and dynamically reflect the human body's health status, emotional changes, and physiological responses to external stimuli. Emotion recognition refers to the process of identifying human emotional states through technological means. It attempts to determine a person's current emotion from their behavior, language, or physiological signals. Since physiological signals are the human body's direct physiological responses to emotional stimuli and are less controlled by subjective consciousness, physiological parameters can provide a more objective and realistic assessment of emotional states, and are especially suitable for scenarios requiring high reliability.

[0003] However, at present, when identifying users' emotions, facial expressions are often used as the basis for identification, rather than physiological parameters such as electrocardiogram and respiratory signals, which leads to a low accuracy rate of emotion recognition. Therefore, this invention proposes an emotion recognition method and system based on physiological parameter extraction. Summary of the Invention

[0004] The purpose of this invention is to propose an emotion recognition method and system based on physiological parameter extraction, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: Firstly, an emotion recognition method based on physiological parameter extraction, the method comprising: Step S1: Set the emotion induction methods corresponding to different emotion categories, and collect multiple sets of electrocardiogram signals of the test individual under different emotion categories; Step S2: Extract features from multiple sets of electrocardiogram (ECG) signals of the individual under test in the time domain and frequency domain respectively to obtain the time domain features and frequency domain features corresponding to the ECG signals; Step S3: Use the time-domain and frequency-domain features corresponding to multiple sets of electrocardiogram signals of the individual to be tested as a training set to construct an emotion classification method; Step S4: Collect the user's electrocardiogram signal within a recent time period and input it into the emotion classification method. Generate the user's corresponding recognized emotion through the emotion classification method and output it.

[0006] Further, step S1 includes the following sub-steps: Step S101: Obtain multiple emotion categories and set corresponding emotion triggering methods based on the emotion categories; Step S102: Set up a corresponding emotional induction scenario using the emotional induction method, bind an electrocardiogram (ECG) signal acquisition device to the individual to be tested, and collect the individual's ECG signal within the emotional induction scenario; Step S103: Identify the actual emotion category felt by the individual in the emotionally induced scenario; Step S104: Compare the actual emotion category with the emotion category corresponding to the emotion-inducing scenario; if the actual emotion category is the same as the emotion category corresponding to the emotion-inducing scenario, then retain the corresponding electrocardiogram signal. If the actual emotion category is different from the emotion category corresponding to the emotion-inducing scenario, the corresponding ECG signal will be discarded, and then the emotion-inducing scenario will be reset until all ECG signals are collected.

[0007] Further, step S2 includes the following sub-steps: Step S201: Obtain multiple sets of electrocardiogram (ECG) signals of the individual to be tested, obtain multiple sets of ECG signals corresponding to the individual to be tested, and identify the ECG waveform corresponding to the ECG signals; Step S202: The voltage change rate of the previous moment is obtained by dividing the difference between the voltages at adjacent moments by the duration of the adjacent moments. Step S203: Identify multiple extreme points in the electrocardiogram waveform, select an electrocardiogram waveform with an extreme point distribution of minimum point-maximum point-minimum point and record it as the pre-selected R wave region; Step S204: Read the voltage change rate on both sides of the maximum point in the preselected R-wave region; The absolute values ​​of the voltage change rates on both sides of the maximum point are compared with the voltage change rate threshold. If the absolute values ​​of the voltage change rates on both sides of the maximum point are greater than or equal to the voltage change rate threshold, the pre-selected R-wave region is recorded as the actual R-wave region. If the absolute value of the voltage change rate on either side of the maximum point is less than the voltage change rate threshold, then the corresponding pre-selected R-wave region is discarded.

[0008] Furthermore, step S2 also includes the following sub-steps: Step S205: The area of ​​change on both sides of the first minimum point in the actual R-wave region is recorded as the Q-wave, and the area of ​​change on both sides of the second minimum point in the R-wave region is recorded as the S-wave; counting forward based on the Q-wave, the area on both sides of the first minimum point is recorded as the P-wave; counting backward based on the S-wave, the area on both sides of the first minimum point is recorded as the T-wave. Step S206: Record the electrocardiogram waveform between one P wave and the next P wave as one heartbeat, and count the duration of the P wave region, the duration of the R wave region and the duration of the S wave region in each heartbeat. Step S207: Identify the time corresponding to the left endpoint of the R wave region in the electrocardiogram waveform. Record the time corresponding to the left endpoint of the first R wave region as the first heart rate time, the time corresponding to the left endpoint of the second R wave region as the second heart rate time, and so on, to obtain the third heart rate time to the Nth heart rate time; N is a constant. Step S208: Subtract the first heart rate time from the second heart rate time to obtain the first heart rate interval; subtract the second heart rate time from the third heart rate time to obtain the second heart rate interval; and so on, to obtain N-1 sets of heart rate intervals corresponding to the electrocardiogram waveform.

[0009] Furthermore, step S2 also includes the following sub-steps: Step S209: Sum the N-1 groups of heart rate intervals, calculate the mean value to obtain the mean value of the heart rate intervals corresponding to multiple groups of heart rate intervals, and calculate the heart rate variability XB corresponding to the electrocardiogram using the formula as follows: Where XLJ is the mean heart rate interval, XLi is the heart rate interval corresponding to the i-th group of the electrocardiogram waveform, and i is the number of the heart rate interval, i=1,2,...,N-1; Step S210: Record the heart rate variability and the duration of the P wave region, the duration of the R wave region and the duration of the S wave region in each heartbeat as the time-domain characteristics of the corresponding electrocardiogram signal. Step S211: Perform a Fourier transform on the electrocardiogram waveform to obtain a frequency domain electrocardiogram waveform.

[0010] Furthermore, step S2 also includes the following sub-steps: Step S212: Divide the frequency domain electrocardiogram waveform according to frequency. The portion of the frequency domain electrocardiogram waveform from zero to the first segmented frequency is recorded as the first electrocardiogram frequency band; the portion of the frequency domain waveform waveform from the first segmented frequency to the second segmented frequency is recorded as the second electrocardiogram frequency band; the portion of the frequency domain waveform waveform from the second segmented frequency to the third segmented frequency is recorded as the third electrocardiogram frequency band; and the portion greater than the third segmented frequency is recorded as the noise frequency band. Step S213: Discard the noise frequency band and calculate the power spectral density of the ECG signal in the first, second, and third ECG frequency bands using the formula. Step S214: The power spectral density of the second ECG frequency band is divided by the power spectral density of the third ECG frequency band to obtain the power spectral ratio value corresponding to the ECG signal. Step S215: The power spectral density of the ECG signal in the first ECG frequency band, the second ECG frequency band, and the third ECG frequency band, and the power spectral density ratio of the ECG signal are recorded as the frequency domain features of the corresponding ECG signal.

[0011] Furthermore, the specific formulas for calculating the power spectral density corresponding to the first, second, and third ECG frequency bands are as follows: The power spectral density YGMj in the first ECG frequency band: YGMj=PP(j) 2 / (PLc / YCD); where PLc is the sampling frequency, YCD is the length of the first ECG frequency band, PP(j) is the expression obtained by Fourier transform of the ECG waveform, specifically the amplitude spectrum; j is the index of the corresponding amplitude spectrum; The power spectral density EGMk of the second ECG frequency band: EGMk=PP(k) 2 / (PLc / ECD); the power spectral density corresponding to the first ECG frequency band, ECD is the length of the second ECG frequency band; PP(k) is the expression obtained by Fourier transform of the ECG waveform, specifically the amplitude spectrum, and k is the index of the amplitude spectrum corresponding to the second ECG frequency band; The power spectral density (TGMh) of the third ECG frequency band: TGMh = PP(h) 2 / (PLc / TCD); the power spectral density corresponding to the first ECG frequency band, TCD is the length of the second ECG frequency band, PP(h) is the expression obtained by Fourier transform of the ECG waveform, specifically the amplitude spectrum, and h is the index of the amplitude spectrum corresponding to the third ECG frequency band.

[0012] Further, step S3 includes the following sub-steps: Step S301: Obtain the time-domain and frequency-domain features corresponding to multiple sets of electrocardiogram signals of the individual to be tested; Step S302: Obtain the relationship table between the actual emotion category and the three-dimensional emotion of the individual to be tested, match the actual emotion category and the three-dimensional emotion relationship table to obtain the identification of emotional state, stress state and fatigue state of ECG signals of different groups; Step S303: Using time-domain features and frequency-domain features as inputs, and sequentially using the identification of emotional state, stress state, and fatigue state as outputs, an emotion classification method is constructed based on the classification algorithm. Step S304: Collect multiple sets of electrocardiogram signals corresponding to the individual to be tested again and record them as verification parameter data. Extract the time domain features and frequency domain features corresponding to the verification parameter data and import them into the emotion classification method to obtain the output emotional state, output stress state and output fatigue state corresponding to the verification parameter data.

[0013] Furthermore, step S3 also includes the following sub-steps: Step S305: Output emotional state, output stress state, and output fatigue state, and compare them sequentially with the corresponding recognized emotional state, recognized stress state, and recognized fatigue state in the verification parameter data; if the output emotional state, output stress state, and output fatigue state are all the same as the corresponding recognized emotional state, recognized stress state, and recognized fatigue state, then the recognition is recorded as successful; otherwise, the recognition is recorded as failed. Step S306: Count the number of successful recognitions, divide the number of successful recognitions by the number of groups of verification parameter data to obtain the recognition success rate; when the recognition success rate is greater than or equal to the success rate threshold, output the emotion classification method; when the recognition success rate is less than the success rate threshold, regenerate the emotion classification method.

[0014] Secondly, an emotion recognition system based on physiological parameter extraction includes a data acquisition module, a time-domain feature extraction module, a frequency-domain feature extraction module, a method training module, and an emotion recognition module. The data acquisition module is used to set the emotion induction mode corresponding to different emotion categories, and to collect multiple sets of electrocardiogram signals of the test individual under different emotion categories and send them to the time domain feature extraction module and the frequency domain feature extraction module; The time-domain feature extraction module is used to extract the time-domain features corresponding to multiple sets of electrocardiogram signals of the test individual under different emotion categories and send them to the method training module; the frequency-domain feature extraction module is used to extract the frequency-domain features corresponding to multiple sets of electrocardiogram signals of the test individual under different emotion categories and send them to the method training module. The training module is used to construct a specific emotion classification method based on the time-domain and frequency-domain features corresponding to multiple sets of electrocardiogram signals of the individual under test as a training set; the emotion recognition module is used to generate and output the corresponding emotion of the user based on the emotion recognition method.

[0015] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. This invention first sets up emotion induction methods corresponding to different emotion categories, and collects multiple sets of electrocardiogram (ECG) signals of the test individual under different emotion categories; it then extracts features from the multiple sets of ECG signals of the test individual in the time domain and frequency domain respectively, and obtains the time domain features and frequency domain features corresponding to the ECG signals; this invention analyzes the ECG status of the test individual and extracts the ECG features of the test individual. 2. This invention uses the time-domain and frequency-domain features corresponding to multiple sets of electrocardiogram (ECG) signals of the individual under test as a training set to construct an emotion classification method; finally, it collects the user's ECG signals in a recent time period and substitutes them into the emotion classification method, generates the user's corresponding recognized emotion through the emotion classification method and outputs it; this invention achieves accurate identification of user emotions through ECG signal extraction. Attached Figure Description

[0016] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0017] Figure 1 This is a flowchart illustrating the overall method of the present invention; Figure 2 This is a schematic diagram of the extreme points corresponding to the central electrical waveform of the present invention; Figure 3 This is a schematic diagram of the central electrical waveform corresponding to different waves in this invention; Figure 4 This is a system structure block diagram of the system involved in this invention. Detailed Implementation

[0018] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: Please refer to Figures 1-3 As shown, the technical solution provided by this invention is as follows: an emotion recognition method based on physiological parameter extraction, which involves collecting multiple sets of electrocardiogram signals from an individual to be tested, and then extracting the frequency domain features and time domain features of the individual to be tested in the time domain and frequency domain, respectively; identifying the corresponding emotional state, stress state, and fatigue state in the time domain and frequency domain based on the frequency domain features and time domain features; training an emotion recognition model by using the frequency domain features and time domain features as inputs and the identified emotional state, stress state, and fatigue state as outputs; and generating and outputting the user's corresponding identified emotion through the emotion recognition model.

[0020] In this embodiment, the emotion recognition method includes: Step S1: Set the emotion induction methods corresponding to different emotion categories, and collect multiple sets of electrocardiogram signals of the test individual under different emotion categories; The specific categories of emotions include: joy, peace, sadness, and anger, etc. In this invention, step S1 includes the following sub-steps: Step S101: Obtain multiple emotion categories and set corresponding emotion induction methods based on the emotion categories; wherein, the emotion induction methods include: material stimulation method, situational recall method, and situational induction method; It should be noted that emotions more easily triggered directly by external stimuli are preferentially induced using the material stimulus method; emotions related to an individual's internal memory and experience are preferentially induced using the situational recall method; and emotions that can only be fully experienced in a specific situation are preferentially induced using the situational induction method. Preferably, the correspondence between emotion induction methods and emotion categories is as follows: The emotional categories corresponding to the stimulus method are: joy, sadness, disgust, or surprise. The emotional categories corresponding to the situational recall method are: calm, anger, fear, or tension. The emotional categories corresponding to the situational evoked method are: focus, tension, fear, or boredom; Specifically, the methods for inducing emotions should be adjusted based on the actual situation of the individual being tested, and will not be disclosed here. Step S102: Set up a corresponding emotional induction scenario using the emotional induction method, bind an electrocardiogram (ECG) signal acquisition device to the individual to be tested, and collect the individual's ECG signal within the emotional induction scenario; Specifically, the emotional triggering scenarios for the material stimulation method can be: joy - playing a piece of cheerful music or a comedy clip, or showing a set of pictures full of joy and smiles; The emotionally eliciting scenario for the situational recall method can be: calm - guide the individual to recall a peaceful afternoon, sitting alone on a park bench, enjoying the breeze and sunshine; The emotional induction method corresponds to the emotional induction scenario: by creating an emotion-related situation, the individual being tested experiences the target emotion in a simulated environment; Step S103: Identify the actual emotion category felt by the individual in the emotionally induced scenario; Specifically, this can be obtained by having the individual fill in a questionnaire about the actual emotional category they felt in the emotionally triggered scenario; Step S104: Compare the actual emotion category with the emotion category corresponding to the emotion-inducing scenario; if the actual emotion category is the same as the emotion category corresponding to the emotion-inducing scenario, then retain the corresponding electrocardiogram signal. If the actual emotion category is different from the emotion category corresponding to the emotion-inducing scenario, the corresponding ECG signal will be discarded, and then the emotion-inducing scenario will be reset until all ECG signals are collected.

[0021] Step S2: Extract features from multiple sets of electrocardiogram (ECG) signals of the individual under test in the time domain and frequency domain respectively to obtain the time domain features and frequency domain features corresponding to the ECG signals; In this invention, step S20 includes the following sub-steps: Step S201: Acquire multiple sets of electrocardiogram (ECG) signals of the individual to be tested, obtaining multiple sets of ECG signals corresponding to the individual to be tested, such as... Figures 2-3As shown, the ECG waveform corresponding to the ECG signal is identified. The horizontal axis of the heart rate waveform represents time, and the vertical axis represents voltage. Step S202: The voltage change rate of the previous moment is obtained by dividing the difference between the voltages at adjacent moments by the duration of the adjacent moments. Step S203: Identify multiple extreme points in the electrocardiogram waveform, select an electrocardiogram waveform with an extreme point distribution of minimum point-maximum point-minimum point and record it as the pre-selected R wave region; Among them, extreme points are divided into maximum points and minimum points. The ordinate of a maximum point is greater than the ordinate of the points to its left and right; the ordinate of a minimum point is less than the ordinate of the points to its left and right. Step S204: Read the voltage change rate on both sides of the maximum point in the preselected R-wave region (i.e., the voltage change rate between the previous and next time moments corresponding to the maximum point); take the absolute value of the voltage change rate on both sides of the maximum point and compare it with the voltage change rate threshold; if the absolute value of the voltage change rate on both sides of the maximum point is greater than or equal to the voltage change rate threshold, then the preselected R-wave region is recorded as the actual R-wave region. If the absolute value of the voltage change rate on either side of the maximum point is less than the voltage change rate threshold, then the corresponding pre-selected R-wave region is discarded. Step S205: The area of ​​change on both sides of the first minimum point in the actual R-wave region is recorded as the Q-wave, and the area of ​​change on both sides of the second minimum point in the R-wave region is recorded as the S-wave; counting forward based on the Q-wave, the area on both sides of the first minimum point is recorded as the P-wave; counting backward based on the S-wave, the area on both sides of the first minimum point is recorded as the T-wave. Step S206: Record the electrocardiogram waveform between one P wave and the next P wave as one heartbeat, and count the duration of the P wave region, the duration of the R wave region and the duration of the S wave region in each heartbeat. Step S207: Identify the time corresponding to the left endpoint of the R wave region in the electrocardiogram waveform. Record the time corresponding to the left endpoint of the first R wave region as the first heart rate time, the time corresponding to the left endpoint of the second R wave region as the second heart rate time, and so on, to obtain the third heart rate time to the Nth heart rate time; N is a constant. Step S208: Subtract the first heart rate time from the second heart rate time to obtain the first heart rate interval; subtract the second heart rate time from the third heart rate time to obtain the second heart rate interval; and so on, to obtain N-1 sets of heart rate intervals corresponding to the electrocardiogram waveform. Step S209: Sum the N-1 groups of heart rate intervals, calculate the mean value to obtain the mean value of the heart rate intervals corresponding to multiple groups of heart rate intervals, and calculate the heart rate variability XB corresponding to the electrocardiogram using the formula as follows: Where XLJ is the mean heart rate interval, XLi is the heart rate interval corresponding to the i-th group of the electrocardiogram waveform, and i is the number of the heart rate interval, i=1,2,...,N-1; It should be noted that heart rate variability is an important indicator reflecting arrhythmia in the individual being tested; Step S210: Record the heart rate variability and the duration of the P wave region, the duration of the R wave region and the duration of the S wave region in each heartbeat as the time-domain characteristics of the corresponding electrocardiogram signal. Step S211: Perform a Fourier transform on the electrocardiogram waveform to obtain a frequency domain electrocardiogram waveform; wherein, the horizontal axis of the frequency domain electrocardiogram waveform is frequency, and the vertical axis of the frequency domain electrocardiogram waveform is amplitude; Step S212: Divide the frequency domain electrocardiogram waveform according to frequency. The portion of the frequency domain electrocardiogram waveform from zero to the first segmented frequency is recorded as the first electrocardiogram frequency band; the portion of the frequency domain waveform waveform from the first segmented frequency to the second segmented frequency is recorded as the second electrocardiogram frequency band; the portion of the frequency domain waveform waveform from the second segmented frequency to the third segmented frequency is recorded as the third electrocardiogram frequency band; and the portion greater than the third segmented frequency is recorded as the noise frequency band. Step S213: Discard the noise frequency band and calculate the power spectral density of the ECG signal in the first, second, and third ECG frequency bands using the following formula: Power spectral density of the first ECG frequency band: YGMj=PP(j) 2 / (PLc / YCD); where PLc is the sampling frequency, YCD is the length of the first ECG frequency band, PP(j) is the expression obtained by Fourier transform of the ECG waveform, specifically the amplitude spectrum; j is the index of the corresponding amplitude spectrum; Specifically, the sampling frequency refers to the number of samples collected per second when discretizing ECG signals over a continuous time period. It is generally taken as 1000Hz and can be considered a constant. The length of the first ECG frequency band is the total number of ECG signals collected. The total time length is calculated by dividing the length of the first ECG frequency band by the sampling frequency. Power spectral density of the second ECG frequency band: EGMk=PP(k) 2 / (PLc / ECD); the power spectral density corresponding to the first ECG frequency band, ECD is the length of the second ECG frequency band; PP(k) is the expression obtained by Fourier transform of the ECG waveform, specifically the amplitude spectrum, and k is the index of the amplitude spectrum corresponding to the second ECG frequency band; Power spectral density of the third ECG band: TGMh = PP(h) 2 / (PLc / TCD); the power spectral density corresponding to the first ECG frequency band, TCD is the length of the second ECG frequency band, PP(h) is the expression obtained by Fourier transform of the ECG waveform, specifically the amplitude spectrum, and h is the index of the amplitude spectrum corresponding to the third ECG frequency band; Step S214: The power spectral density of the second ECG frequency band is divided by the power spectral density of the third ECG frequency band to obtain the power spectral ratio value corresponding to the ECG signal. Step S215: The power spectral density of the ECG signal in the first ECG frequency band, the second ECG frequency band, and the third ECG frequency band, and the power spectral density ratio of the ECG signal are recorded as the frequency domain features of the corresponding ECG signal.

[0022] Step S3: Use the time-domain and frequency-domain features corresponding to multiple sets of electrocardiogram signals of the individual to be tested as a training set to construct an emotion classification method; In this invention, step S3 includes the following sub-steps: Step S301: Obtain the time-domain and frequency-domain features corresponding to multiple sets of electrocardiogram signals of the individual to be tested; Step S302: Obtain the relationship table between the actual emotion category and the three-dimensional emotion of the individual to be tested, match the actual emotion category and the three-dimensional emotion relationship table to obtain the identification of emotional state, stress state and fatigue state of ECG signals of different groups; Emotional Three-Dimensional Relationship Table:

[0023] Step S303: Using time-domain features and frequency-domain features as inputs, and sequentially using the identification of emotional state, stress state, and fatigue state as outputs, a unique emotion classification method is constructed based on the classification algorithm. Classification algorithms are tools in machine learning and data mining. Their core principle is to analyze sample data of known categories and build a model that can automatically classify new data into predefined categories. Optional classification algorithms include decision trees, logistic regression, neural networks, and Naive Bayes. Taking the decision tree algorithm as an example: the information gain ratio corresponding to all time-domain features and frequency-domain features is calculated using the ID3 algorithm; where information gain ratio is a feature selection metric in the decision tree algorithm, used to measure the degree of contribution of an attribute to the classification result; Arrange them in descending order of information gain rate, and then select the corresponding time domain features or frequency domain features for classification in turn, and take the corresponding features as root nodes (such as frequency domain features of electrocardiogram signals). A decision tree is constructed by setting a splitting threshold at the root node. The data at the root node is compared to the corresponding threshold. If the data is greater than or equal to the threshold, it moves to the left tree; otherwise, it moves to the right tree. For example: The power spectral density and the power spectral ratio of the ECG signal corresponding to the first and third ECG frequency bands are both greater than the corresponding thresholds; the power spectral density and the power spectral ratio of the ECG signal corresponding to the second ECG frequency band are less than the corresponding thresholds. In this case, the power spectral density and the power spectral ratio of the ECG signal corresponding to the first and third ECG frequency bands are entered into the left tree, and the power spectral ratio of the ECG signal corresponding to the second ECG frequency band is entered into the right tree. Recursive calculations are performed in the left tree until the node purity reaches a threshold (e.g., 90% of samples belong to the same category) or the maximum depth (e.g., 5 layers) is reached, at which point the current decision tree is considered complete; this process is repeated to construct multiple decision trees. The corresponding feature data are sequentially input into the decision tree. Multiple decision trees are combined to obtain the output emotional state, output stress state, and output fatigue state, and then output them to realize the construction of an emotion classification method. Step S304: Collect multiple sets of ECG signals corresponding to the individual to be tested again and record them as verification parameter data. Extract the time domain features and frequency domain features corresponding to the verification parameter data and import them into the emotion classification method to obtain the output emotional state, output stress state and output fatigue state corresponding to the verification parameter data. Step S305: Output emotional state, output stress state, and output fatigue state, and compare them sequentially with the corresponding recognized emotional state, recognized stress state, and recognized fatigue state in the verification parameter data; if the output emotional state, output stress state, and output fatigue state are all the same as the corresponding recognized emotional state, recognized stress state, and recognized fatigue state, then the recognition is recorded as successful; otherwise, the recognition is recorded as failed. Step S306: Count the number of successful recognitions, divide the number of successful recognitions by the number of groups of verification parameter data to obtain the recognition success rate; when the recognition success rate is greater than or equal to the success rate threshold, output the emotion classification method; when the recognition success rate is less than the success rate threshold, regenerate the emotion classification method.

[0024] Step S4: Collect the user's electrocardiogram signal within a recent time period and input it into the emotion classification method. The emotion classification method generates the user's corresponding recognized emotion and outputs it.

[0025] Example 2, as a supplement to Example 1, can add step S5 between step S2 and step S3, which extracts the time-domain and frequency-domain features corresponding to electromyography (EMG), skin point signals, and respiratory electrical signals, and performs emotion recognition based on EMG, skin point signals, and respiratory electrical signals assisted by ECG signals. In this embodiment, step S5 involves extracting features from the electromyography (EMG), electrodermal (ED) signals, and respiratory electrical signals of the individual under test in both the time and frequency domains; wherein the physiological parameter data includes EMG signals, EED signals, and respiratory electrical signals. Step S5 includes the following sub-steps: Step S501: Obtain multiple sets of physiological parameter data of the individual to be tested, and obtain multiple sets of electromyography signals, electrodermal signals and respiratory signals corresponding to the individual to be tested; Step S502: Convert the electromyographic signal into an electromyographic waveform and identify the electromyographic voltage value JD at different times in the waveform. i Set the extraction window duration T, and divide the electromyography waveform based on the extraction window duration to obtain multiple electromyography segmentation windows; where i is the number of different times in the extraction window; Step S503: The electromyographic voltage values ​​at different times within the same electromyographic segmentation window are summed and averaged to obtain the mean electromyographic value CJZx of the corresponding electromyographic segmentation window; where x is the number of the different electromyographic segmentation window. Step S504: Calculate the mean square value of electromyography (EMG) of the window corresponding to the EMG segmentation window, CJFx, the first-order difference of EMG in the window, CYCx, and the second-order difference of EMG in the window, CECx, using the formulas as follows: Mean square value of electromyography at window: ; First-order difference of window electromyography: ; Second-order difference of window electromyography: ; Step S505: Calculate the mean value, mean square value, first-order difference, and second-order difference of the electromyography (EMG) signals corresponding to all EMG segmentation windows, and summarize the mean value, mean square value, first-order difference, and second-order difference of the EMG signals as the temporal features of the corresponding EMG signals. Step S506: Perform the calculation process of steps S211-S215 to obtain the frequency domain features corresponding to the electromyographic signal. Step S507: Perform the process described in steps S302-S306 on the electrodermal signal and the electrical respiratory signal to obtain the time domain features and frequency domain features corresponding to the electrodermal signal and the electrical respiratory signal; It should be noted that electromyography (EMG), electrical skin signal, and respiratory signal are all bioelectrical signals. Therefore, the basic principle of calculating the frequency domain characteristics of EMG signals, as well as the time domain and frequency domain characteristics of electrical skin and respiratory signals, is the same. However, in specific calculations, appropriate fine-tuning should be made according to the signal characteristics; for example, the sampling frequency should be distinguished.

[0026] Example 3, as Figure 4As shown, based on another concept of the same invention, an emotion recognition system based on physiological parameter extraction is proposed, including a data acquisition module, a time-domain feature extraction module, a frequency-domain feature extraction module, a method training module, and an emotion recognition module: The data acquisition module is used to set the emotion induction mode corresponding to different emotion categories, and to collect multiple sets of physiological parameter data of the test individual under different emotion categories and send them to the time domain feature extraction module and the frequency domain feature extraction module; wherein, the physiological parameter data includes: electrocardiogram signal, electromyogram signal, skin conductance signal and respiratory conductance signal; The temporal feature extraction module is used to extract the temporal features corresponding to multiple sets of physiological parameter data of the individual under different emotion categories and send them to the method training module. The frequency domain feature extraction module is used to extract the frequency domain features corresponding to multiple sets of physiological parameter data of the individual under different emotion categories and send them to the method training module. The training module of the method is used to construct a unique emotion classification method based on the time-domain and frequency-domain features corresponding to multiple sets of physiological parameter data of the individual under test as a training set. The emotion recognition module is used to generate and output the user's corresponding emotion based on the emotion recognition method.

[0027] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An emotion recognition method based on physiological parameter extraction, characterized in that, The methods include: Step S1: Set the emotion induction methods corresponding to different emotion categories, and collect multiple sets of electrocardiogram signals of the test individual under different emotion categories; Step S2: Extract features from multiple sets of electrocardiogram (ECG) signals of the individual under test in the time domain and frequency domain respectively to obtain the time domain features and frequency domain features corresponding to the ECG signals; Step S3: Use the time-domain and frequency-domain features corresponding to multiple sets of electrocardiogram signals of the individual to be tested as a training set to construct an emotion classification method; Step S4: Collect the user's electrocardiogram signal within a recent time period and input it into the emotion classification method. Generate the user's corresponding recognized emotion through the emotion classification method and output it.

2. The emotion recognition method based on physiological parameter extraction according to claim 1, characterized in that, Step S1 includes the following sub-steps: Step S101: Obtain multiple emotion categories and set corresponding emotion triggering methods based on the emotion categories; Step S102: Set up a corresponding emotional induction scenario using the emotional induction method, bind an electrocardiogram (ECG) signal acquisition device to the individual to be tested, and collect the individual's ECG signal within the emotional induction scenario; Step S103: Identify the actual emotion category felt by the individual in the emotionally induced scenario; Step S104: Compare the actual emotion category with the emotion category corresponding to the emotion-inducing scenario; If the actual emotion category is the same as the emotion category corresponding to the emotion-inducing scenario, the corresponding ECG signal will be retained. If the actual emotion category is different from the emotion category corresponding to the emotion-inducing scenario, the corresponding ECG signal will be discarded, and then the emotion-inducing scenario will be reset until all ECG signals are collected.

3. The emotion recognition method based on physiological parameter extraction according to claim 1, characterized in that, Step S2 includes the following sub-steps: Step S201: Obtain multiple sets of electrocardiogram (ECG) signals of the individual to be tested, obtain multiple sets of ECG signals corresponding to the individual to be tested, and identify the ECG waveform corresponding to the ECG signals; Step S202: The voltage change rate of the previous moment is obtained by dividing the difference between the voltages at adjacent moments by the duration of the adjacent moments. Step S203: Identify multiple extreme points in the electrocardiogram waveform, select an electrocardiogram waveform with an extreme point distribution of minimum point-maximum point-minimum point and record it as the pre-selected R wave region; Step S204: Read the voltage change rate on both sides of the maximum point in the preselected R-wave region; The absolute values ​​of the voltage change rates on both sides of the maximum point are compared with the voltage change rate threshold. If the absolute values ​​of the voltage change rates on both sides of the maximum point are greater than or equal to the voltage change rate threshold, the pre-selected R-wave region is recorded as the actual R-wave region. If the absolute value of the voltage change rate on either side of the maximum point is less than the voltage change rate threshold, then the corresponding pre-selected R-wave region is discarded.

4. The emotion recognition method based on physiological parameter extraction according to claim 3, characterized in that, Step S2 further includes the following sub-steps: Step S205: The area of ​​change on both sides of the first minimum point in the actual R-wave region is recorded as the Q-wave, and the area of ​​change on both sides of the second minimum point in the R-wave region is recorded as the S-wave; counting forward based on the Q-wave, the area on both sides of the first minimum point is recorded as the P-wave; counting backward based on the S-wave, the area on both sides of the first minimum point is recorded as the T-wave. Step S206: Record the electrocardiogram waveform between one P wave and the next P wave as one heartbeat, and count the duration of the P wave region, the duration of the R wave region and the duration of the S wave region in each heartbeat. Step S207: Identify the time corresponding to the left endpoint of the R wave region in the electrocardiogram waveform. Record the time corresponding to the left endpoint of the first R wave region as the first heart rate time, the time corresponding to the left endpoint of the second R wave region as the second heart rate time, and so on, to obtain the third heart rate time to the Nth heart rate time; N is a constant. Step S208: Subtract the first heart rate time from the second heart rate time to obtain the first heart rate interval; subtract the second heart rate time from the third heart rate time to obtain the second heart rate interval; and so on, to obtain N-1 sets of heart rate intervals corresponding to the electrocardiogram waveform.

5. The emotion recognition method based on physiological parameter extraction according to claim 4, characterized in that, Step S2 further includes the following sub-steps: Step S209: Sum the N-1 groups of heart rate intervals, calculate the mean value to obtain the mean value of the heart rate intervals corresponding to multiple groups of heart rate intervals, and calculate the heart rate variability XB corresponding to the electrocardiogram using the formula as follows: Where XLJ is the mean heart rate interval, XLi is the heart rate interval corresponding to the i-th group of the electrocardiogram waveform, and i is the number of the heart rate interval, i=1,2,...,N-1; Step S210: Record the heart rate variability and the duration of the P wave region, the duration of the R wave region and the duration of the S wave region in each heartbeat as the time-domain characteristics of the corresponding electrocardiogram signal. Step S211: Perform a Fourier transform on the electrocardiogram waveform to obtain a frequency domain electrocardiogram waveform.

6. The emotion recognition method based on physiological parameter extraction according to claim 5, characterized in that, Step S2 further includes the following sub-steps: Step S212: Divide the frequency domain electrocardiogram waveform according to frequency. The part of the frequency domain electrocardiogram waveform from zero to the first segmented frequency is recorded as the first electrocardiogram frequency band; the part of the frequency domain waveform waveform from the first segmented frequency to the second segmented frequency is recorded as the second electrocardiogram frequency band. The portion of the frequency domain waveform from the second segmented frequency to the third segmented frequency is designated as the third ECG frequency band; the portion above the third segmented frequency is designated as the noise frequency band. Step S213: Discard the noise frequency band and calculate the power spectral density of the ECG signal in the first, second, and third ECG frequency bands using the formula. Step S214: The power spectral density of the second ECG frequency band is divided by the power spectral density of the third ECG frequency band to obtain the power spectral ratio value corresponding to the ECG signal. Step S215: The power spectral density of the ECG signal in the first ECG frequency band, the second ECG frequency band, and the third ECG frequency band, and the power spectral density ratio of the ECG signal are recorded as the frequency domain features of the corresponding ECG signal.

7. The emotion recognition method based on physiological parameter extraction according to claim 6, characterized in that, The specific formulas for calculating the power spectral density corresponding to the first, second, and third ECG frequency bands are as follows: The power spectral density YGMj in the first ECG frequency band: YGMj=PP(j) 2 / (PLc / YCD); where PLc is the sampling frequency, YCD is the length of the first ECG frequency band, PP(j) is the expression obtained by Fourier transform of the ECG waveform, specifically the amplitude spectrum; j is the index of the corresponding amplitude spectrum; The power spectral density EGMk of the second ECG frequency band: EGMk=PP(k) 2 / (PLc / ECD); the power spectral density corresponding to the first ECG frequency band, ECD is the length of the second ECG frequency band; PP(k) is the expression obtained by Fourier transform of the ECG waveform, specifically the amplitude spectrum, and k is the index of the amplitude spectrum corresponding to the second ECG frequency band; The power spectral density (TGMh) of the third ECG band: TGMh = PP(h) 2 / (PLc / TCD); the power spectral density corresponding to the first ECG frequency band, TCD is the length of the second ECG frequency band, PP(h) is the expression obtained by Fourier transform of the ECG waveform, specifically the amplitude spectrum, and h is the index of the amplitude spectrum corresponding to the third ECG frequency band.

8. The emotion recognition method based on physiological parameter extraction according to claim 1, characterized in that, Step S3 includes the following sub-steps: Step S301: Obtain the time-domain and frequency-domain features corresponding to multiple sets of electrocardiogram signals of the individual to be tested; Step S302: Obtain the relationship table between the actual emotion category and the three-dimensional emotion of the individual to be tested, match the actual emotion category and the three-dimensional emotion relationship table to obtain the identification of emotional state, stress state and fatigue state of ECG signals of different groups; Step S303: Using time-domain features and frequency-domain features as inputs, and sequentially using the identification of emotional state, stress state, and fatigue state as outputs, an emotion classification method is constructed based on the classification algorithm. Step S304: Collect multiple sets of electrocardiogram signals corresponding to the individual to be tested again and record them as verification parameter data. Extract the time domain features and frequency domain features corresponding to the verification parameter data and import them into the emotion classification method to obtain the output emotional state, output stress state and output fatigue state corresponding to the verification parameter data.

9. The emotion recognition method based on physiological parameter extraction according to claim 8, characterized in that, Step S3 further includes the following sub-steps: Step S305: Output emotional state, output stress state, and output fatigue state, and compare them sequentially with the corresponding recognized emotional state, recognized stress state, and recognized fatigue state in the verification parameter data; if the output emotional state, output stress state, and output fatigue state are all the same as the corresponding recognized emotional state, recognized stress state, and recognized fatigue state, then the recognition is recorded as successful; otherwise, the recognition is recorded as failed. Step S306: Count the number of successful recognitions, divide the number of successful recognitions by the number of groups of verification parameter data to obtain the recognition success rate; when the recognition success rate is greater than or equal to the success rate threshold, output the emotion classification method; when the recognition success rate is less than the success rate threshold, regenerate the emotion classification method.

10. An emotion recognition system based on physiological parameter extraction, characterized in that, An emotion recognition method based on physiological parameter extraction according to any one of claims 1-9 includes a data acquisition module, a time-domain feature extraction module, a frequency-domain feature extraction module, a method training module, and an emotion recognition module. The data acquisition module is used to set the emotion induction mode corresponding to different emotion categories, and to collect multiple sets of electrocardiogram signals of the test individual under different emotion categories and send them to the time domain feature extraction module and the frequency domain feature extraction module; The time-domain feature extraction module is used to extract the time-domain features corresponding to multiple sets of electrocardiogram signals of the test individual under different emotion categories and send them to the method training module; the frequency-domain feature extraction module is used to extract the frequency-domain features corresponding to multiple sets of electrocardiogram signals of the test individual under different emotion categories and send them to the method training module. The training module of the method is used to construct a specific emotion classification method based on the time-domain and frequency-domain features corresponding to multiple sets of electrocardiogram signals of the individual under test as a training set. The emotion recognition module is used to generate and output the user's corresponding emotion based on the emotion recognition method.

Images