Emotional state detection system based on electrophysiological data

Abstract

The invention relates to the technical field of biological information processing, in particular to an emotional state detection system based on electrophysiological data, which comprises a signal receiving and calibrating module, a signal anomaly detection module, an emotional mode analysis module and an emotional output module. According to the invention, through accurate calibration and time synchronization of electrocardiogram, brain wave and skin electrical activity signals, the availability and data accuracy of the signals are improved, the reliability of emotional state detection is ensured, abnormal indexes are accurately eliminated by adopting an advanced signal processing technology, the purity of the signals is improved, and the accuracy of emotional state detection is improved. A solid foundation is provided for accurate analysis of emotional characteristics, various physiological signals are comprehensively analyzed, the system is allowed to respond to subtle emotional changes in real time, and the precision of emotional detection is improved by adjusting signal processing parameters and optimizing system adaptability and response speed based on emotional stability indexes. And the system is more efficient and sensitive in application in mental health monitoring and human-computer interaction.

Description

Technical Field

[0001] The present invention relates to the technical field of biological information processing, and in particular to an emotional state detection system based on electrophysiological data. Background Art

[0002] The field of bioinformatics processing technology focuses on the use of computational methods to process and analyze biological data. The main tasks of this field include collecting, storing, analyzing and integrating biological data to provide in-depth analysis of biological systems. It has a wide range of applications, from genomics and proteomics to complex biological systems and bioelectric signal analysis. Bioinformatics processing technology enables scientists to analyze batches of biological data, mine current biomarkers, design drugs, and better analyze the basis of genetic diseases. A key application in this field is to extract useful information from bioelectric signals through pattern recognition and signal processing technology to support clinical diagnosis and biomedical research.

[0003] Among them, the emotional state detection system is a technology that identifies emotional states based on monitoring and analyzing individual physiological signals. This system uses physiological indicators such as heart rate, skin electrical activity, and brain waves to parse data through special algorithms to infer the individual's emotional state, such as happiness, sadness, anger, etc. The main uses of the emotional state detection system include applications in mental health monitoring, user experience research, and augmented reality technology, to assist in improving the human-computer interaction experience, and to provide users with more personalized services and products that respond to their emotional needs.

[0004] Existing technologies face problems of inconsistent signal quality and synchronization difficulties when processing and analyzing bioelectric signals, especially in the case of multiple signal sources, such as the fusion processing of electrocardiograms, brain waves and skin electrical activity signals, which leads to inaccurate data analysis and affects the accurate identification of emotional states. Traditional technologies lack effective mechanisms for processing abnormal signals and are difficult to accurately remove noise and outliers from complex biological signals, limiting the effectiveness and universal adaptability of emotion detection systems in practical applications. For example, if abnormal signals are not effectively removed, misjudgments will occur in mental health monitoring, and in user experience research, erroneous data will lead to incorrect interpretations of user emotions, limiting the application of bioinformatics processing technology in a wider range of fields. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings in the prior art and to propose an emotional state detection system based on electrophysiological data.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme: an emotional state detection system based on electrophysiological data comprises:

[0007] The signal receiving and calibration module receives the signals of electrocardiogram, brain wave and skin electrical activity, performs signal calibration and corrects the time deviation to obtain the calibrated signal, and synchronizes the electrocardiogram waveform, brain wave pattern and skin electrical response according to the calibrated signal to obtain the synchronized electrophysiological data;

[0008] The signal anomaly detection module uses a convolutional neural network to identify arrhythmias and abnormal brain waves based on the synchronized electrophysiological data, obtains abnormal indicators, removes and corrects the abnormal indicators, and generates purified electrophysiological signals;

[0009] The emotion pattern analysis module uses the purified electrophysiological signal to analyze the electrocardiogram waveform and the electroencephalogram pattern through an autoregressive moving average model to obtain preliminary emotion characteristics, and performs emotion labeling through the preliminary emotion characteristics to generate emotion classification results;

[0010] The emotion output module uses the emotion classification result, receives emotion tagging data, calculates the frequency and intensity of emotion changes in differentiated time periods, determines the statistical characteristics of emotion fluctuations, obtains an emotion stability index, adjusts signal processing parameters according to the emotion stability index, optimizes signal analysis accuracy, and generates adjusted emotion recognition results.

[0011] As a further solution of the present invention, the step of acquiring the synchronous electrophysiological data is specifically as follows:

[0012] By receiving the signals of electrocardiogram, brain wave and skin electrical activity, executing the time calibration algorithm, the time deviation caused by the transmission delay is eliminated, and the signal set after time calibration is obtained;

[0013] The time-calibrated signal set is synchronized using the formula:

[0014]

[0015] Calculate the difference of each physiological signal and generate a synchronized signal set, where x i,a represents the sample value of the i-th signal, represents the average value of the signal samples, n represents the number of samples, S a Indicates the degree of synchronization;

[0016] The synchronized signal set is utilized to optimize signal processing by dynamically adjusting synchronization parameters to obtain synchronized electrophysiological data.

[0017] As a further solution of the present invention, the step of obtaining the abnormal indicator is specifically as follows:

[0018] Extracting features of electrocardiogram and brain waves from the synchronized electrophysiological data, including peaks, troughs and signal strengths, and extracting feature information using target digital signal processing technology to obtain a physiological signal feature set;

[0019] Based on the physiological signal feature set, arrhythmia and abnormal brain waves are identified using the formula:

[0020] I b =σ(W b ·F b +b b )

[0021] Through the propagation of the network, the abnormal pattern data is obtained, where σ represents the activation function and W b represents the weight of the convolutional layer, b b Indicates bias, I b Indicates abnormal mode data, F b represents a physiological signal feature set;

[0022] The abnormal pattern data is analyzed, the frequency and severity of the differentiated abnormal patterns are calculated, and the abnormal indicators are obtained according to the preset critical thresholds.

[0023] As a further solution of the present invention, the step of obtaining the purified electrophysiological signal is specifically as follows:

[0024] Based on the abnormal indicators, analyzing the ECG and EEG signals, identifying the locations of arrhythmias and abnormal EEG waves, and generating an abnormal segment list;

[0025] Apply a filter to the abnormal section list to process the data, fine-tune the signal, reduce noise interference, and use the formula:

[0026] R c =α c ·L c -β c ∑(γ c ·L c +δ c )

[0027] Output the corrected signal data set, where α c , β c , γ c and δ c are parameters adjusted according to the abnormal signal strength and frequency, R c represents the corrected signal data set, L c Indicates the abnormal segment of the mark;

[0028] The corrected signal data set is evaluated, statistical analysis is performed and signal quality is verified, frequency distribution detection is used to verify the physiological credibility and integrity of the signal, and a purified electrophysiological signal is generated.

[0029] As a further solution of the present invention, the step of obtaining the preliminary emotion characteristics is specifically as follows:

[0030] By performing frequency domain analysis on the purified electrophysiological signal, converting the signal from the time domain to the frequency domain using Fourier transform, calculating the frequency spectrum density of the electrocardiogram and electroencephalogram signals, and generating frequency spectrum density features of the electrocardiogram and electroencephalogram;

[0031] According to the spectral density characteristics of the ECG and EEG, the autoregressive moving average model is applied, combined with the waveform data of the ECG and EEG signals, and the formula is adopted:

[0032]

[0033] Calculate the electrophysiological signal characteristics associated with emotions and obtain the autoregressive and moving average parameters of the emotional state, where α d and β d Represent the autoregressive and moving average coefficients, y j-d Represents the original ECG or EEG signal value of the jth signal, ∈ j-d represents the original prediction error, m represents the number of samples, A d represents the autoregressive and moving average parameters of the emotional state;

[0034] The emotional state autoregression and moving average parameters are used to perform threshold analysis and identify emotional changes, set an emotional detection threshold, and compare whether the prediction error of each calculation cycle exceeds the threshold to verify the change of the emotional state and obtain preliminary emotional characteristics.

[0035] As a further solution of the present invention, the step of obtaining the emotion classification result is specifically as follows:

[0036] By using the preliminary emotional features, according to the changes in the electrocardiogram and electroencephalogram signals, a corresponding emotional state is assigned to each time point to obtain labeled time series data;

[0037] The labeled time series data is used as input, a weighted calculation method is applied and multiple emotion features are combined, and the formula is adopted:

[0038]

[0039] Calculate the total sentiment score and get the sentiment score sequence, where E h represents the total sentiment score, w k,h represents the weight of the emotion feature k, F k,h is the standardized score of emotion feature k, bh is the bias term, K represents the total emotional feature quantity;

[0040] Based on the emotion score sequence, the support vector machine algorithm is used to map the continuous emotion scores to emotion categories, including happiness, sadness, and anger, and the emotion state at each time point is classified to generate emotion classification results.

[0041] As a further solution of the present invention, the step of obtaining the emotional stability index is specifically as follows:

[0042] By analyzing the emotion tag data of the emotion classification result, an index is established for the emotion change in each time period, and the emotion data is converted into a time series format to generate an initialized emotion data time series;

[0043] Based on the initialized emotional data time series, calculate the emotional fluctuation characteristics within the differentiated time period, determine the frequency and intensity of emotional changes, and use the formula:

[0044]

[0045] Get emotional fluctuation characteristic data, where F g represents the contribution of emotional fluctuations, r t,g represents the frequency of emotion change at time point t, e t,g is the emotional intensity at time point t, b g is the bias term, T represents the total time points;

[0046] Based on the emotion fluctuation characteristic data, by combining the standard deviation and coefficient of variation of emotion fluctuation, the overall stability of emotion is evaluated, the predictability of the emotion state within the measurement period is measured, and the emotion stability index is generated.

[0047] As a further solution of the present invention, the step of obtaining the adjusted emotion recognition result is specifically:

[0048] Based on the emotional stability index, it is evaluated whether the current signal processing parameters are adapted to the current emotional index. By comparing the current emotional index with a predetermined threshold, if the index shows a large deviation, the parameter adjustment process is triggered, using the formula:

[0049] P u =α u ×ISI u +β u

[0050] Generate preliminary adjustment signal processing parameters, where ISI u represents the emotional stability index, α u is the coefficient of sensitivity to emotion change, β u is the baseline correction factor, Pu represents the initial adjustment of signal processing parameters;

[0051] The signal processing parameters adjusted in the preliminary manner are applied to signal analysis, simulation tests are performed and the impact on emotion recognition accuracy is evaluated by performing multiple simulations and applying the formula:

[0052] E u =γ u ×P 2 u +δ u

[0053] Estimate recognition accuracy, verify the effect of new parameter settings, and generate simulation test results, where γ u represents the adjustment factor, δ u is a constant offset, E u represents the simulation test results, P u represents the initial adjustment of signal processing parameters;

[0054] The signal processing parameters are adjusted and solidified according to the simulation test results, the advantages of the new parameters are verified by comparing the emotion recognition results before and after the adjustment, and the adjusted emotion recognition results are output.

[0055] Compared with the prior art, the advantages and positive effects of the present invention are:

[0056] In the present invention, through the careful calibration and time synchronization processing of electrocardiogram, brain wave and skin electrical activity signals, the availability and data accuracy of the signals are significantly improved, and the basic data quality of subsequent analysis is ensured, thereby enhancing the reliability of emotional state detection. In abnormality detection, abnormal indicators are accurately eliminated through advanced signal processing technology, which not only improves the purity of the signal, but also lays a solid foundation for the accurate analysis of emotional characteristics. Through comprehensive reference to multiple physiological signals, a deeper analysis of complex emotional states is achieved, allowing real-time response to subtle changes in emotions. Through the application of the emotional stability index, the signal processing parameters are adjusted, and the adaptability and response speed of the system are further optimized, which effectively improves the accuracy and response speed of the emotion detection system, making its application in mental health monitoring and human-computer interaction experience more efficient and sensitive. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is a system flow chart of the present invention;

[0058] Figure 2 is a flow chart of synchronizing electrophysiological data in the present invention;

[0059] Figure 3 It is a flow chart of abnormal indicators in the present invention;

[0060] Figure 4is a flow chart of the electrophysiological signal after purification in the present invention;

[0061] Figure 5 is a flow chart of preliminary emotion features in the present invention;

[0062] Figure 6 A flow chart of the emotion classification results in the present invention;

[0063] Figure 7 is a flow chart of the emotional stability index in the present invention;

[0064] Figure 8 Flow chart of the adjusted emotion recognition result in the present invention. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0066] In the description of the present invention, it is necessary to interpret that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, which are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be interpreted as a limitation on the present invention. In addition, in the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0067] See also Figure 1 , the emotional state detection system based on electrophysiological data includes:

[0068] The signal receiving and calibration module receives the signals of electrocardiogram, brain wave and skin electrical activity, performs signal calibration and corrects the time deviation to obtain the calibrated signal. According to the calibrated signal, the electrocardiogram waveform, brain wave pattern and skin electrical response are synchronized to obtain the synchronized electrophysiological data.

[0069] The signal anomaly detection module uses convolutional neural networks to identify arrhythmias and abnormal brain waves based on synchronized electrophysiological data, obtains abnormal indicators, removes and corrects the abnormal indicators, and generates purified electrophysiological signals;

[0070] The emotion pattern analysis module uses the purified electrophysiological signals to analyze the ECG waveform and EEG pattern through the autoregressive moving average model to obtain preliminary emotion features, and then labels the emotions based on the preliminary emotion features to generate emotion classification results.

[0071] The emotion output module uses the emotion classification results, receives emotion tagging data, calculates the frequency and intensity of emotion changes in differentiated time periods, determines the statistical characteristics of emotion fluctuations, obtains the emotion stability index, adjusts the signal processing parameters according to the emotion stability index, optimizes the signal analysis accuracy, and generates adjusted emotion recognition results.

[0072] The calibrated signals include heart rate synchronization data, brain wave frequency matching data, and skin electrodermal response time alignment data. The synchronized electrophysiological data include aligned electrocardiogram, standardized electroencephalogram, and synchronized electrodermagraph. Abnormal indicators include arrhythmia types and abnormal brain wave bands. The purified electrophysiological signals include denoised electrocardiogram signals, filtered electroencephalogram signals, and corrected skin electrodermal signals. Preliminary emotional features include electrocardiogram emotion indicators and electroencephalogram emotion patterns. The emotion classification results include the identified basic emotion types and emotional response levels. The emotion stability index includes the frequency of emotion fluctuations and the measurement of emotion response intensity. The adjusted emotion recognition results include the optimized classification accuracy and parameter adjustment details.

[0073] See also Figure 2 ,The specific steps for acquiring synchronous electrophysiological data are:

[0074] By receiving the signals of electrocardiogram, brain wave and skin electrical activity, executing the time calibration algorithm, the time deviation caused by the transmission delay is eliminated, and the signal set after time calibration is obtained;

[0075] Based on the original electrocardiogram, brain wave, and skin electrodermal activity data collected by the sensor, a time calibration algorithm is executed to eliminate the time deviation caused by transmission delay. The collected raw data undergoes a preprocessing stage, including filtering and denoising, to ensure the clarity and availability of the data. The filters used in this stage are customized according to the characteristics of the signal. For example, electrocardiogram data uses a bandpass filter to eliminate high-frequency and low-frequency noise, while brain wave and skin electrodermal activity data require differentiated types of filters to process. Each set of data is time synchronized after calibration to ensure signal alignment in subsequent processing stages. Time calibration uses the network time protocol to ensure the alignment of timestamps on all devices, involving complex signal processing techniques such as signal resampling and time interpolation. The processing ensures the accuracy and synchronization of the data before entering the analysis stage, and obtains a time-calibrated signal set.

[0076] The time-calibrated signal set is synchronized using the formula:

[0077]

[0078] Calculate the difference of each physiological signal and generate a synchronized signal set, where x i,a represents the sample value of the i-th signal, represents the average value of the signal samples, n represents the number of samples, S a Indicates the degree of synchronization;

[0079] The benefit of the formula is that it provides a method to quantify the synchronization error, which can effectively evaluate the accuracy of the synchronization processing of differentiated physiological signals;

[0080] S a It represents the degree of synchronization, which is a quantitative indicator used to measure the accuracy of synchronization between multiple physiological signals (such as electrocardiogram, brain waves, and skin electrical activity);

[0081] Given n=3 (representing three signals: ECG, EEG, and skin electricity), let the ECG signal value x 1,a =0.45mv, EEG signal value x 2,a =0.48mv, skin electrical signal value x 3,a =0.47mv, the average value of all signals Substitute the values ​​into the formula to calculate the degree of synchronization:

[0082]

[0083] The result shows that the synchronization error is very small, indicating that the signal synchronization processing has achieved high precision, providing a reliable data basis for the next step of analysis.

[0084] Using the synchronized signal set, the signal processing is optimized by dynamically adjusting the synchronization parameters to obtain synchronized electrophysiological data;

[0085] Using the synchronized signal set generated in the previous step, the signal processing is optimized by dynamically adjusting the synchronization parameters to ensure the synchronization accuracy of the final output physiological signal. The dynamic adjustment involves calculating the statistical characteristics of each signal, such as the mean and standard deviation, and adjusting the parameters of the synchronization algorithm according to the characteristics, such as adjusting the time window or resampling rate. The process is iterative, and the synchronization accuracy is evaluated at each iteration, and the parameters are adjusted according to the evaluation results until the preset accuracy standard is reached, and finally the synchronized electrophysiological data is obtained.

[0086] See also Figure 3 , the specific steps for obtaining abnormal indicators are:

[0087] Extract the features of electrocardiogram and brain waves from the synchronized electrophysiological data, including peaks, troughs and signal strength, and use the target digital signal processing technology to extract the feature information to obtain the physiological signal feature set;

[0088] Extracting features of the ECG and brain waves from synchronized electrophysiological data involves precisely measuring the timing and amplitude of each waveform to ensure that the most critical features are extracted from the physiological signals, such as the R wave peak and Q wave trough of the ECG, and the alpha and beta waves of the brain waves. The features reflect the basic patterns of heart and brain activity and are processed through specially designed filters and signal processing algorithms to ensure that the extracted features are free of noise interference. The feature data obtained through processing is the key basis for identifying abnormal physiological states. The data provides the necessary input for subsequent machine learning models for more in-depth analysis and anomaly detection.

[0089] Based on the physiological signal feature set, arrhythmia and abnormal brain waves are identified using the formula:

[0090] I b =σ(W b ·F b +b b )

[0091] Through the propagation of the network, the abnormal pattern data is obtained, where σ represents the activation function and W b represents the weight of the convolutional layer, b b Indicates bias, I b Indicates abnormal mode data, F b represents a physiological signal feature set;

[0092] The benefit of the formula is that through the convolution operation, the patterns of arrhythmia and abnormal brain waves can be effectively extracted from the ECG and EEG features. The convolution layer can capture the local dependencies and spatial hierarchical structures in the feature data, which enhances the model's ability to detect abnormal physiological signals.

[0093] Set W b The weight matrix is ​​[0.5, -0.2, 0.8], and the bias b b is 0.1, the activation function σ uses ReLU, and the input feature F b is [1, 0, 1], the calculation process is as follows:

[0094] W b ·F b +b b =0.5×1+(-0.2)×0+0.8×1+0.1=1.4

[0095] I b =σ(1.4)=max(0,1.4)=1.4

[0096] The results show that under the given weights and features, the network successfully identifies the abnormal pattern and outputs 1.4, indicating the strength of the detected abnormality. The larger the value, the more obvious the abnormality, which provides an important basis for clinical diagnosis.

[0097] Analyze abnormal pattern data, calculate the frequency and severity of differentiated abnormal patterns, and obtain abnormal indicators based on preset critical thresholds;

[0098] Analyze the abnormal pattern data identified by the convolutional neural network, refer to the clinical significance of the differentiated abnormal patterns, calculate the statistical indicators of the patterns, such as frequency and duration, determine which abnormal patterns are temporary or sporadic, and indicate long-term health problems, judge the severity of the abnormality by the set critical threshold, classify each abnormal state, and integrate the analysis results into the final abnormality indicators, which describe in detail the type, frequency and severity of the detected arrhythmias and abnormal brain waves.

[0099] See also Figure 4 , the steps for obtaining the purified electrophysiological signal are as follows:

[0100] Based on abnormal indicators, analyze ECG and EEG signals, identify the locations of arrhythmias and abnormal brain waves, and generate a list of abnormal segments;

[0101] For the abnormal segments identified in the synchronous electrophysiological signals, a detailed time point analysis is performed. According to the specific frequency and amplitude of the irregular patterns in the ECG and EEG signals, the appropriate filtering technology is selected to accurately remove interference. Advanced signal processing technologies such as fast Fourier transform (FFT) are used to analyze the frequency domain characteristics of the abnormal segments, so as to locate the exact start and end time points. The precise positioning of the abnormal segments in the signal is the key basis for the subsequent correction processing. Through this process, a detailed list of abnormal segments can be generated. This list will serve as the basis for the next cleaning and correction operations, laying a solid foundation for achieving high-precision signal correction.

[0102] Apply filters to the abnormal segment list to process the data, fine-tune the signal, reduce noise interference, and use the formula:

[0103] R c =α c ·L c -β c ∑(γ c ·L c +δ c )

[0104] Output the corrected signal data set, where α c , β c , γ c and δ c are parameters adjusted according to the abnormal signal strength and frequency, R c represents the corrected signal data set, L c Indicates the abnormal segment of the mark;

[0105] The benefit of the formula is that by adjusting four key parameters (α c , β c , γ c , δ c ), which allows flexible control of the contribution of each component in the signal correction process, thereby improving the correction accuracy and reducing non-critical signal distortion. It is especially suitable for processing physiological signals with high variability. This formula combines linear amplification and nonlinear combination to optimize the processing of abnormal signal segments;

[0106] The following parameter values ​​are set:

[0107] α c =0.95, representing the adjustment coefficient of the overall signal, which determines the degree of signal amplification or attenuation;

[0108] β c =0.1, determines the contribution of the abnormal signal component to the final output signal;

[0109] γ c =1.5, adjust the weight of a single abnormal signal item;

[0110] δ c = 0.05, the correction term adds a constant offset to deal with baseline drift;

[0111] L c Represents a typical abnormal signal set, including three signal segments: L c ={0.2,-0.3,0.15};

[0112] The calculation process is as follows:

[0113] R c =0.95·{0.2,-0.3,0.15}-0.1·∑(1.5·{0.2,-0.3,0.15}+0.05)

[0114] Calculate each term:

[0115] ={0.19,-0.285,0.1425}-0.1

[0116] ·{(1.5·0.2+0.05),(1.5·-0.3+0.05),(1.5·0.15+0.05)}

[0117] ={0.19,-0.285,0.1425}-0.1·{0.35,-0.4,0.275}

[0118] ={0.19,-0.285,0.1425}-{0.035,-0.04,0.0275}

[0119] ={0.155,-0.245,0.115}

[0120] The results show that after adjustment by the above formula, the abnormal part of the original signal has been effectively corrected, and each signal segment has been specifically adjusted to ensure the purification and quality improvement of the entire signal, so that the corrected signal R c More suitable for further analysis and diagnosis.

[0121] Evaluate the corrected signal data set, perform statistical analysis and verify signal quality, use frequency distribution detection to verify the physiological credibility and integrity of the signal, and generate cleaned electrophysiological signals;

[0122] The signals that have been deeply cleaned and corrected need to be quality evaluated. The signals are passed through a series of statistical analysis procedures to test their signal-to-noise ratio and fidelity, ensuring that all processed signals do not lose key information for clinical diagnosis. Frequency distribution tests and complexity analysis are used to evaluate the overall structure and functionality of the signals to ensure their effectiveness and reliability in medical research and diagnostic applications. Through a series of evaluation steps, purified electrophysiological signals are generated. The signals integrate the results of all previous steps and provide high-quality data resources for further medical analysis.

[0123] See also Figure 5 ,The steps to obtain the preliminary emotional features are as follows:

[0124] By performing frequency domain analysis on the purified electrophysiological signals, converting the signals from the time domain to the frequency domain using Fourier transform, calculating the spectrum density of the ECG and EEG signals, and generating the spectrum density features of the ECG and EEG signals;

[0125] The purified electrophysiological signals are subjected to fast Fourier transform, which involves converting the time series signals into frequency domain signals and analyzing the frequency characteristics of the ECG and EEG signals. Frequency analysis can reveal the main components and noise levels in the signals, thereby judging the clarity and availability of the signals. By calculating the frequency distribution, the variability and stability of the heart rate and brain waves can be further analyzed. This process requires window function processing of the time series data and calculation of the Fourier transform of each window to obtain the spectral density distribution during the entire recording period and generate the spectral density characteristics of the ECG and EEG.

[0126] According to the spectral density characteristics of ECG and EEG, the autoregressive moving average model is applied, combined with the waveform data of ECG and EEG signals, and the formula is adopted:

[0127]

[0128] Calculate the electrophysiological signal characteristics associated with emotions and obtain the autoregressive and moving average parameters of the emotional state, where α d and β d Represent the autoregressive and moving average coefficients, y j-d Represents the original ECG or EEG signal value of the jth signal, ∈ j-d represents the original prediction error, m represents the number of samples, A d represents the autoregressive and moving average parameters of the emotional state;

[0129] The benefit of the formula is that it combines the autoregressive coefficient and the moving average coefficient to predict emotional changes. By adding the square term, the sensitivity to extreme values ​​is enhanced, which helps to more accurately identify abnormal changes in ECG and EEG signals when emotions fluctuate.

[0130] Consider a simplified example where the sample size m = 10 and the autoregressive coefficient α d =0.5, moving average coefficient β d =0.3, original ECG or EEG signal value y j-d Take 0.8, the original prediction error ∈ j-d Take 0.2 and substitute it into the formula:

[0131]

[0132] The results show that under the given autoregressive and moving average parameters, the prediction error eigenvalue of the emotional state is 0.332, indicating that the emotional state of the ECG and EEG signals has a moderate degree of fluctuation, which is helpful for further analysis and processing of ECG and EEG signals to identify emotional changes.

[0133] Using the emotional state autoregression and moving average parameters, threshold analysis is performed to identify emotional changes, the emotional detection threshold is set, and the prediction error of each calculation cycle is compared to see if it exceeds the threshold, to verify the change of emotional state and obtain preliminary emotional characteristics;

[0134] Using the calculated emotion-related electrophysiological features, a threshold is set to identify significant emotion changes. The process involves comparing the prediction error of the calculation cycle with the set threshold. This threshold is determined based on statistical analysis of historical data to ensure that emotional fluctuations that are inconsistent with the norm can be effectively identified. After setting the threshold, significant emotion changes are detected by comparing the error value cycle by cycle to see if it exceeds the threshold. If the prediction error exceeds the threshold, it is marked as a significant emotion change. This method can not only dynamically track changes in emotional states, but also take corresponding response measures in a timely manner when emotions change significantly, ensuring the accuracy and real-time nature of emotion analysis. The determined emotion changes are marked as preliminary emotion features.

[0135] See also Figure 6 ,The specific steps for obtaining the emotion classification results are:

[0136] Through preliminary emotional features, according to the changes in ECG and EEG signals, the corresponding emotional state is assigned to each time point to obtain labeled time series data;

[0137] The initial step is to obtain preliminary emotion change markers from the previous analysis. This marker is based on significant changes in ECG and EEG signals. Through advanced signal processing techniques such as filtering and feature extraction, the signal at each time point is analyzed to determine whether there is a significant emotion change. This involves batch data collection and preprocessing, including denoising and normalization, to ensure data quality and consistency of processing. Each valid emotion change is marked and converted into a series of marked time series data that can be used for further analysis. The data will be used as input to the emotion recognition model to provide basic data for subsequent emotion classification.

[0138] Taking the labeled time series data as input, applying the weighted calculation method and combining multiple emotion features, the formula is adopted:

[0139]

[0140] Calculate the total sentiment score and get the sentiment score sequence, where E h represents the total sentiment score, w k,h represents the weight of the emotion feature k, F k,h is the standardized score of emotion feature k, b h is the bias term, K represents the total emotional feature quantity;

[0141] The benefit of the formula is that by adding a bias term to adjust the baseline of the emotion score, the model is allowed to remain flexible under different individuals and situations, thereby improving the accuracy and adaptability of overall emotion recognition;

[0142] There are 3 emotional features, with weights w 1,h =0.5, w 2,h =0.3, w 3,h =0.2, the corresponding emotional feature score is F 1,h =70, F 2,h =60, F 3,h =50, the bias term is b h =5, the calculation process is:

[0143] E h =(0.5·70+0.3·60+0.2·50+5)=(35+18+10+5)=68

[0144] The result shows that the comprehensive emotion score is 68, which means that under the given weights and features, the identified emotional state is a moderate emotional response, and this score will be used for further classification judgment.

[0145] Based on the emotion score sequence, the support vector machine algorithm is used to map the continuous emotion scores to emotion categories, including happiness, sadness, and anger, and the emotion state at each time point is classified to generate emotion classification results;

[0146] Classification is performed using emotion score sequences. By adopting classification algorithms in machine learning, such as support vector machines or decision trees, the algorithm learns how to map emotion scores to specific emotion categories through training data. The classification process includes model training, verification, and testing to ensure the accuracy and reliability of the classification results. By analyzing and processing differentiated emotion scores, the algorithm can identify specific emotional states, such as happiness, sadness, or anger. The emotional state at each time point is accurately classified to provide end users with intuitive emotion recognition results.

[0147] See also Figure 7 , the specific steps for obtaining the emotional stability index are:

[0148] By analyzing the emotion tag data of the emotion classification results, an index is established for the emotion changes in each time period, and the emotion data is converted into a time series format to generate an initialized emotion data time series;

[0149] The analysis process begins by extracting emotional marker data at key time points from the emotional classification results. This involves screening data associated with specific time points from batches of emotional records, and performing time series analysis on the extracted data in order to conduct an in-depth study of the dynamic changes in emotions. First, each marker is time-sorted to ensure the coherence and logical correctness of the analysis. By comparing the changes in emotional intensity between each consecutive time point, the frequency of emotional fluctuations is calculated. The processing process not only refers to the absolute value of emotional intensity, but also includes the speed and trend of emotional intensity changes. Through the refined steps, the pattern of emotional changes can be accurately captured, laying a solid data foundation for the next step of in-depth analysis, generating a series of time series emotional data ready for further analysis. The data will be directly used to evaluate the statistical characteristics of emotional fluctuations, ensuring the high quality of data processing and the reliability of analysis results.

[0150] Based on the initialized emotional data time series, calculate the emotional fluctuation characteristics within the differentiated time period, determine the frequency and intensity of emotional changes, and use the formula:

[0151]

[0152] Get emotional fluctuation characteristic data, where F grepresents the contribution of emotional fluctuations, r t,g represents the frequency of emotion change at time point t, e t,g is the emotional intensity at time point t, b g is the bias term, T represents the total time points;

[0153] The formula is beneficial in that it refers to the frequency and intensity of emotional changes through time series weighting, which improves the flexibility and accuracy of the analysis of emotional change characteristics;

[0154] Suppose that the emotional data is monitored for T = 30 days, where the frequency of emotional changes per day is r t,g and emotional intensity t,g Obtained from actual monitoring, for example, on a certain day, 15,g =0.8, e 15,g =0.5, set the bias term b g =0.1, then the emotional fluctuation contribution of this day is F g =0.8×0.5+0.1=0.5, the total emotional fluctuation characteristic of the entire time period is calculated cumulatively for all days, which can accurately quantify the emotional changes of each day or even every moment, and accumulate their impact on the overall emotional stability. This calculation method not only scientifically reflects the actual changes in emotions, but also can assist psychologists and researchers to better analyze the nature and influencing factors of emotional fluctuations. The result shows that the frequency and intensity of emotional changes can be comprehensively referenced through formula calculation to reflect emotional stability.

[0155] Based on the data of emotional fluctuation characteristics, by combining the standard deviation and coefficient of variation of emotional fluctuations, the overall stability of emotions is evaluated, the predictability of emotional states within the measurement cycle is measured, and an emotional stability index is generated;

[0156] In the final stage of obtaining the emotional stability index, we deeply analyze the statistical characteristics of emotional data. We use statistical methods to calculate the basic statistics of the data, such as the mean, median, standard deviation, and coefficient of variation, through the sorted time series emotional data. Statistical indicators help us analyze the prevalence and abnormality of emotional fluctuations. The standard deviation provides a measure of the amplitude of emotional intensity fluctuations, while the coefficient of variation reflects the volatility of emotional intensity relative to the average level. Through quantitative indicators, we can reveal not only the intensity of emotional changes, but also the regularity and predictability of their fluctuations. By comparing emotional data in different time periods (such as weekdays and weekends), we can identify the specific impact of life events on emotional stability. Detailed analysis will accumulate into key data for calculating the emotional stability index. The index can comprehensively reflect the individual's emotional fluctuations and mental health status, providing a scientific basis for subsequent psychological intervention and emotional management.

[0157] See also Figure 8, the steps to obtain the adjusted emotion recognition results are as follows:

[0158] Based on the emotional stability index, evaluate whether the current signal processing parameters are suitable for the current emotional index. By comparing the current emotional index with the predetermined threshold, if the index shows a large deviation, the parameter adjustment process is triggered, using the formula:

[0159] P u =α u ×ISI u +β u

[0160] Generate preliminary adjustment signal processing parameters, where ISI u represents the emotional stability index, α u is the coefficient of sensitivity to emotion change, β u is the baseline correction factor, P u represents the initial adjustment of signal processing parameters;

[0161] The benefit of the formula is that by combining the emotional stability index and adjustment parameters, this formula enables the signal processing parameters to more accurately reflect the user's emotional changes, optimizes the accuracy of emotion recognition, and is based on the latest emotional stability index ISI. u , this formula adjusts the signal processing parameters to capture emotional fluctuations more delicately, where α u reflects the sensitivity of emotional changes to signal processing, while β u is the baseline adjustment factor, which ensures the stability of the system when there is no significant change in emotion;

[0162] To verify the signal processing parameter P u To test the effectiveness of the adjustment, a series of simulation tests were conducted. First, the simulated emotional stability index (ISI) was set. u =0.55, according to the previously set adjustment parameter α u = 2.0 and the baseline adjustment factor β u =0.6, calculate the current signal processing parameters: P u =2.0×0.55+0.6=1.7;

[0163] Using this new parameter P u Signal parsing simulations were conducted in a controlled environment, including signal inputs with differentiated emotion intensities, to evaluate the accuracy and response time of emotion recognition before and after parameter adjustment.

[0164] The initially adjusted signal processing parameters were applied to signal parsing, simulation tests were conducted and the impact on emotion recognition accuracy was evaluated. By performing multiple simulations, the formula was applied:

[0165] E u =γ u×P 2 u +δ u

[0166] Estimate recognition accuracy, verify the effect of new parameter settings, and generate simulation test results, where γ u represents the adjustment factor, δ u is a constant offset, E u represents the simulation test results, P u represents the initial adjustment of signal processing parameters;

[0167] In the simulation test, using formula E u =γ u ×P 2 u +δ u , to predict the accuracy of emotion recognition, where γ u =0.05,δ u = 0.1, substituting the adjusted signal processing parameters, we can calculate E u =0.05×(1.7) 2 +0.1=0.1445+0.1=0.2445;

[0168] This result represents the expected performance of the adjusted system in the simulated emotion recognition task. u Value, evaluate the actual effect of parameter adjustment on improving emotion recognition accuracy;

[0169] Experimental data show that after adjusting the parameters, the system has significantly improved recognition accuracy when dealing with complex emotional changes, reduced misrecognition rate, and optimized the overall performance of the system. Through this method, the necessity and effectiveness of parameter adjustment are verified, ensuring that each adjustment is based on sufficient experimental verification and data support, which greatly improves the system's sensitivity to emotional changes and recognition accuracy.

[0170] Adjust and solidify the signal processing parameters according to the simulation test results, verify the advantages of the new parameters by comparing the emotion recognition results before and after the adjustment, and output the adjusted emotion recognition results;

[0171] The ultimate goal of the adjustment is to solidify the current signal processing parameters into the emotion recognition system to achieve long-term stable operation. This step involves the ultimate verification, that is, to adjust the parameters P uIt is formally applied in the actual emotion recognition environment and carries out continuous performance monitoring. The monitoring process includes real-time data capture, logging of emotion recognition results, error analysis, etc. It can evaluate the performance of new parameters in actual applications in real time and adjust parameters in time to adapt to the emotional change patterns of differentiated users. This not only improves the accuracy of emotion recognition, but also optimizes the user experience, making the system more humane and meeting the needs of a wider range of users.

[0172] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. An emotional state detection system based on electrophysiological data, characterized in that: The system comprises: The signal receiving and calibration module receives the signals of electrocardiogram, brain wave and skin electrical activity, performs signal calibration and corrects the time deviation to obtain the calibrated signal, and synchronizes the electrocardiogram waveform, brain wave pattern and skin electrical response according to the calibrated signal to obtain the synchronized electrophysiological data; The signal anomaly detection module uses a convolutional neural network to identify arrhythmias and abnormal brain waves based on the synchronized electrophysiological data, obtains abnormal indicators, removes and corrects the abnormal indicators, and generates purified electrophysiological signals; The emotion pattern analysis module uses the purified electrophysiological signal to analyze the electrocardiogram waveform and the electroencephalogram pattern through an autoregressive moving average model to obtain preliminary emotion characteristics, and performs emotion labeling through the preliminary emotion characteristics to generate emotion classification results; The emotion output module uses the emotion classification result, receives emotion tagging data, calculates the frequency and intensity of emotion changes in differentiated time periods, determines the statistical characteristics of emotion fluctuations, obtains an emotion stability index, adjusts signal processing parameters according to the emotion stability index, optimizes signal analysis accuracy, and generates adjusted emotion recognition results.

2. The emotional state detection system based on electrophysiological data according to claim 1, characterized in that: The steps for acquiring the synchronous electrophysiological data are specifically as follows: By receiving the signals of electrocardiogram, brain wave and skin electrical activity, executing the time calibration algorithm, the time deviation caused by the transmission delay is eliminated, and the signal set after time calibration is obtained; The time-calibrated signal set is synchronized using the formula: Calculate the difference of each physiological signal and generate a synchronized signal set, where x i,a represents the sample value of the i-th signal, represents the average value of the signal samples, n represents the number of samples, S a Indicates the degree of synchronization; The synchronized signal set is utilized to optimize signal processing by dynamically adjusting synchronization parameters to obtain synchronized electrophysiological data.

3. The emotional state detection system based on electrophysiological data according to claim 2, characterized in that: The steps for obtaining the abnormal indicators are specifically as follows: Extracting features of electrocardiogram and brain waves from the synchronized electrophysiological data, including peaks, troughs and signal strengths, and extracting feature information using target digital signal processing technology to obtain a physiological signal feature set; Based on the physiological signal feature set, arrhythmia and abnormal brain waves are identified using the formula: I b =σ(W b ·F b +b b ) Through the propagation of the network, the abnormal pattern data is obtained, where σ represents the activation function and W b represents the weight of the convolutional layer, b b Indicates bias, I b Indicates abnormal mode data, F b represents a physiological signal feature set; The abnormal pattern data is analyzed, the frequency and severity of the differentiated abnormal patterns are calculated, and the abnormal indicators are obtained according to the preset critical thresholds.

4. The emotional state detection system based on electrophysiological data according to claim 3, characterized in that: The steps of obtaining the purified electrophysiological signal are specifically as follows: Based on the abnormal indicators, analyzing the ECG and EEG signals, identifying the locations of arrhythmias and abnormal EEG waves, and generating an abnormal segment list; Apply a filter to the abnormal section list to process the data, fine-tune the signal, reduce noise interference, and use the formula: R c =a c ·L c -b c ∑(γ c ·L c +d c ) Output the corrected signal data set, where α c , β c , γ c and δ c are parameters adjusted according to the abnormal signal strength and frequency, R c represents the corrected signal data set, L c Indicates the abnormal segment of the mark; The corrected signal data set is evaluated, statistical analysis is performed and signal quality is verified, frequency distribution detection is used to verify the physiological credibility and integrity of the signal, and a purified electrophysiological signal is generated.

5. The emotional state detection system based on electrophysiological data according to claim 4, characterized in that: The steps for obtaining the preliminary emotion features are specifically as follows: By performing frequency domain analysis on the purified electrophysiological signal, converting the signal from the time domain to the frequency domain using Fourier transform, calculating the frequency spectrum density of the electrocardiogram and electroencephalogram signals, and generating frequency spectrum density features of the electrocardiogram and electroencephalogram; According to the spectral density characteristics of the ECG and EEG, the autoregressive moving average model is applied, combined with the waveform data of the ECG and EEG signals, and the formula is adopted: Calculate the electrophysiological signal characteristics associated with emotions and obtain the autoregressive and moving average parameters of the emotional state, where α d and β d Represent the autoregressive and moving average coefficients, y j-d Represents the original ECG or EEG signal value of the jth signal, ∈ j-d represents the original prediction error, m represents the number of samples, A d represents the autoregressive and moving average parameters of the emotional state; The emotional state autoregression and moving average parameters are used to perform threshold analysis and identify emotional changes, set an emotional detection threshold, and compare whether the prediction error of each calculation cycle exceeds the threshold to verify the change of the emotional state and obtain preliminary emotional characteristics.

6. The emotional state detection system based on electrophysiological data according to claim 5, characterized in that: The steps for obtaining the emotion classification results are specifically as follows: By using the preliminary emotional features, according to the changes in the electrocardiogram and electroencephalogram signals, a corresponding emotional state is assigned to each time point to obtain labeled time series data; The labeled time series data is used as input, a weighted calculation method is applied and multiple emotion features are combined, and the formula is adopted: Calculate the total sentiment score and get the sentiment score sequence, where E h represents the total sentiment score, w k,h represents the weight of the emotion feature k, F k,h is the standardized score of emotion feature k, b h is the bias term, K represents the total emotional feature quantity; Based on the emotion score sequence, the support vector machine algorithm is used to map the continuous emotion scores to emotion categories, including happiness, sadness, and anger, and the emotion state at each time point is classified to generate emotion classification results.

7. The emotional state detection system based on electrophysiological data according to claim 6, characterized in that: The steps for obtaining the emotional stability index are specifically as follows: By analyzing the emotion tag data of the emotion classification result, an index is established for the emotion change in each time period, and the emotion data is converted into a time series format to generate an initialized emotion data time series; Based on the initialized emotional data time series, calculate the emotional fluctuation characteristics within the differentiated time period, determine the frequency and intensity of emotional changes, and use the formula: Get emotional fluctuation characteristic data, where F g represents the contribution of emotional fluctuations, r t,g represents the frequency of emotion change at time point t, e t,g is the emotional intensity at time point t, b g is the bias term, T represents the total time points; Based on the emotion fluctuation characteristic data, by combining the standard deviation and coefficient of variation of emotion fluctuation, the overall stability of emotion is evaluated, the predictability of the emotion state within the measurement period is measured, and the emotion stability index is generated.

8. The emotional state detection system based on electrophysiological data according to claim 7, characterized in that: The steps for obtaining the adjusted emotion recognition result are specifically as follows: Based on the emotional stability index, it is evaluated whether the current signal processing parameters are adapted to the current emotional index. By comparing the current emotional index with a predetermined threshold, if the index shows a large deviation, the parameter adjustment process is triggered, using the formula: P u =a u ×ISI u +b u Generate preliminary adjustment signal processing parameters, where ISI u represents the emotional stability index, α u is the coefficient of sensitivity to emotion change, β u is the baseline correction factor, P u represents the initial adjustment of signal processing parameters; The signal processing parameters adjusted in the preliminary manner are applied to signal analysis, simulation tests are performed and the impact on emotion recognition accuracy is evaluated by performing multiple simulations and applying the formula: Estimate recognition accuracy, verify the effect of new parameter settings, and generate simulation test results, where γ u represents the adjustment factor, δ u is a constant offset, E u represents the simulation test results, P u represents the initial adjustment of signal processing parameters; The signal processing parameters are adjusted and solidified according to the simulation test results, the advantages of the new parameters are verified by comparing the emotion recognition results before and after the adjustment, and the adjusted emotion recognition results are output.

Images