183 results about "Biomedical signal" patented technology

The invention discloses a pressure sensitivity classification evaluation method based on heart rate variability and wearable equipment, and relates to the technical field of biomedical signal processing. According to the method, the collected signals can be processed to form the RR interval sequence and the resultant acceleration, the motion state is judged, the motion state mark is set, the RR interval sequence and the motion state mark are aligned according to the timestamp, the structured data are generated, the pressure sensitivity level of an individual is judged based on the structured data, and the pressure sensitivity level of the individual is calculated. The subjective influence of pressure sensitivity evaluation can be avoided, and the classification precision and adaptability are improved.

The invention discloses a sleep classification method and system based on time-frequency combination, and belongs to the technical field of biomedical signal processing and artificial intelligence. In order to solve the problems that an existing method is high in manual dependence, insufficient in time-frequency feature fusion and low in long-time-sequence modeling efficiency, sleep stage classification is carried out mainly through automatic time-frequency feature extraction, a dynamic attention mechanism with memory enhancement and a lightweight multi-branch neural network structure. According to the method, efficient modeling and accurate classification of the multi-scale electroencephalogram signals can be achieved, the deep sleep recognition capability and the overall classification accuracy are improved, meanwhile, the model parameter quantity and the reasoning delay are remarkably reduced, and good real-time performance and clinical applicability are achieved.

The invention discloses an integrated pediatric vital sign intelligent monitoring and early warning method and system, and relates to the technical field of biomedical signal processing. The method is used for solving the problems of large monitoring interference and early warning lag in a dynamic environment. The method comprises the following steps: firstly, synchronously collecting a multi-modal signal to generate a motion intensity envelope signal and an R wave peak value interval sequence, carrying out orthogonal projection denoising on a photoelectric volume pulse wave signal by utilizing motion envelope based on a recursive least square algorithm, and outputting a reconstructed blood oxygen waveform; then, calculating a cross correlation coefficient of a heart rate trend and a motion envelope, quantifying a physiological coupling degree to generate an independent heart rate anomaly index, and accurately identifying a non-motion related risk; finally, the real-time blood oxygen value is analyzed, the index is used for dynamically increasing the low oxygen threshold value so as to tighten the alarm boundary, physiological parameter linkage grading early warning is achieved, and the recognition sensitivity of the compensation period risk is improved while interference resistance is guaranteed.

The invention discloses a biomedical multi-mode signal anomaly detection and prediction method and system, and the method comprises the steps: modeling a biomedical multi-mode signal into a time-varying non-local dynamic system, capturing the long-time-history dependence characteristic of the signal through a memory mechanism of a Caputo fractional derivative, and introducing a time-varying input item to process interference; the method comprises the following steps of: carrying out high-precision numerical integration by adopting an Adams-Bashform-Module solver; optimizing model parameters in combination with a local domain normalization pre-training strategy and a fusion loss function; abnormal detection is realized by calculating comparison between signal reconstruction deviation and a self-adaptive threshold value; a future anomaly probability is generated based on the trajectory prediction. According to the method, the problems of insufficient non-local dependence capture, poor time-varying interference robustness, high false positive rate and the like in the prior art are effectively solved, the accuracy and real-time performance of abnormal detection and prediction of biomedical signals such as electroencephalogram and electrocardiogram are remarkably improved, and the method is suitable for wearable medical equipment and clinical monitoring systems.

An apparatus for generating diagnostic hypotheses based on electrocardiogram (ECG) data, comprising a processor and a memory containing instructions configuring the processor to generate, using a generative model trained on a corpus, a set of diagnostic hypotheses, wherein generating the set of diagnostic hypotheses includes creating labels, each represents a diagnostic feature associated with diagnostic hypotheses, receive a biomedical signal, identify a biomedical feature as a function of the biomedical signal, select a diagnostic hypothesis from the set of diagnostic hypotheses by matching the biomedical feature against the diagnostic feature, query, as a function of at least a matched label, a medical repository to validate the diagnostic hypothesis, wherein the medical repository includes patients' electronic health records (EHRs), and output the diagnostic hypothesis upon a positive validation of the diagnostic hypothesis.

The invention relates to the technical field of biomedical signal processing, and particularly discloses a fetal growth and development detection method and system based on intellectualization, and the method comprises the steps: obtaining a mixed physiological signal and a fetal movement signal, obtaining a fetal movement interference coefficient and an interference peak period through cross-correlation analysis, constructing an artifact monitoring vector, and predicting a signal deterioration probability; artifact separation is carried out; distinguishing physiological and non-physiological types of the artifact signal, and extracting a signal degradation event; and if the degradation event exists, analyzing a conduction relationship between the event and equipment sampling parameters through a rule engine, generating a main conduction path, constructing an artifact strategy mapping model, and outputting a parameter correction rule. The system comprises a fetal movement analysis module, an artifact separation module, a degradation analysis module and an equipment correction module. According to the method, the multi-dimensional feature analysis and the intelligent model are combined, so that the fetal physiological signal purity and the anomaly detection accuracy can be improved.

The invention discloses a motion posture recognition detection method based on transfer learning, belongs to the field of biomedical signal processing and machine learning, and aims at improving the robustness of a model under the condition of electrode position change by constructing an adversarial training mechanism and aligning data distribution before and after electrode offset in a feature space. The problem of data distortion caused by electrode displacement in the electrical impedance tomography EIT technology is solved, and the accuracy of posture recognition is improved.

The invention discloses a brain heuristic multi-expert multi-mode emotion recognition method and system, equipment and a medium, and belongs to the technical field of artificial intelligence and biomedical signal processing. The method comprises the following steps: by simulating a brain function partitioning mechanism, dividing an electroencephalogram signal into a plurality of brain regions according to neuroanatomy prior, and designing a special expert network for each region; a global-local double-current encoder is adopted to cooperatively extract spatial-temporal characteristics of each brain region signal, and meanwhile, a multi-scale large-kernel convolution module is utilized to extract peripheral physiological signal characteristics; and finally, dynamically fusing multi-expert features through an adaptive routing network to realize sentiment classification. Expert load balancing and bifurcation regularization joint loss are introduced into the model in training, and effective cooperation and feature diversity of experts are ensured. According to the method, excellent recognition precision is obtained in practice, it is verified through interpretability analysis that the decision-making process conforms to neuroscience cognition, and a high-precision and high-reliability solution is provided for application of brain-computer interfaces, mental health monitoring and the like.

The invention belongs to the technical field of biomedical signal processing, and particularly relates to a single-channel pulse signal enhancement method fusing angular domain features and time sequence information, which comprises the following steps: S1, collecting wrist pulse signals by using a single-finger pulse diagnosis bionic hand provided with a pressure sensor; s2, an improved Teager energy operator and a self-adaptive weighted fusion strategy are used for conducting denoising on the pulse signals; s3, adopting an improved Euclidean distance weighted Gramer angle field method to carry out Gramer field conversion on the pulse signals; and S4, using an improved attention score weighting and local standard deviation adjusting method to carry out fusion enhancement on the GASF graph and the GADF graph. According to the method, the denoising performance of the pulse signals is effectively improved through the improved Teager energy operator and the self-adaptive weighted fusion strategy, the time domain dynamic characteristics and structural direction information of the fused image are reserved through the improved Euclidean distance weighted Gramb angle field and the improved GASF and GADF fusion method, and the denoising performance of the pulse signals can be effectively improved through the improved Euclidean distance weighted Gramb angle field and the improved GASF and GADF fusion method. Therefore, a structured enhanced representation mode with angular domain and time sequence information is constructed.

The invention relates to the technical field of audio signal processing and biomedical signal analysis, and still has a further optimized space for the recognition of anti-noise requirements, signal individual differences and complex pathological modes in a noise environment. The invention provides a double-path CNN heart sound classification method based on time-frequency and double-spectrum fusion features, and the method comprises the steps: carrying out the preprocessing of an original heart sound signal of a data set which is classified into a normal heart sound and an abnormal heart sound, and obtaining a to-be-recognized heart sound signal; based on dynamic continuous wavelet transform, adaptively selecting parameters to extract time-frequency characteristics, introducing bispectrum analysis, capturing nonlinear characteristics, generating a dual-channel characteristic pattern, and efficiently storing the dual-channel characteristic pattern in an HDF5 format; and based on a designed double-path convolutional neural network structure, respectively processing the extracted time-frequency and double-spectrum features, performing classification after fusion, and training a model in combination with category weighted loss and an optimization strategy to obtain a heart sound classification result. The heart sound recognition accuracy can be improved.

The invention provides an electroencephalogram abnormal signal detection method based on step-by-step identification and multi-agent decision, and relates to the technical field of biomedical signal processing.The method comprises the steps that electroencephalogram signal data of an epilepsy patient are collected, and the data are analyzed according to a preset channel sequence and cut into a plurality of signal segments with the same length; extracting a multi-dimensional feature from each signal segment; inputting the time domain feature, the frequency domain feature and the inter-channel synchronization feature of each signal segment into an isolated forest model, and screening based on an abnormal proportion threshold to obtain at least one potential abnormal segment; based on the depth scattering feature, the wavelet transform feature, the time domain feature and the frequency domain feature of each potential abnormal segment, executing a multi-agent integration decision on each potential abnormal segment to obtain a comprehensive abnormal score corresponding to each potential abnormal segment; and determining an abnormal signal in the electroencephalogram signal data based on the comprehensive abnormal score corresponding to each potential abnormal segment. According to the method, the false alarm rate of electroencephalogram abnormal signal detection is remarkably reduced.

The invention provides a fetal hypoxia early risk identification method and system based on a large pulse pressure phenotype, and relates to the technical field of biomedical signal processing and heart sound signal analyse.The method comprises the steps that fetal heart sound signals and body surface vibration signals at the abdominal wall of a pregnant woman are collected through a multi-channel sensing array; a preprocessing means is adopted to improve the signal-to-noise ratio of a target signal and suppress an interference signal; in the sliding time window, a large pulse pressure phenotypic feature set and heart sound time phase parameters are extracted; based on the combined criterion, fetal hypoxia risk judgment is carried out; and outputting the graded early warning information, and supporting the playback of the risk event fragment. According to the method, fetal real heart sound mechanical vibration serves as an information source, hypoxia early warning is achieved based on a large pulse pressure phenotype, individualized base lines (including day and night layering) and gestational week self-adaptive correction are supported, the anti-interference capacity is high, the 7 * 24-hour continuous monitoring requirement in a hospital / family can be met, and the sensitivity and specificity of hypoxia recognition are improved.

The invention discloses an emotional state recognition method, system and device, and relates to the technical field of biomedical signal processing and artificial intelligence emotional computation.The method comprises the steps that electroencephalogram signals of a to-be-detected individual are collected; extracting power spectral density (PSD) and differential entropy (DE) features of the electroencephalogram signals on each frequency band according to a preset frequency; performing structured combination on the PSD and DE features corresponding to each frequency band, and constructing a dual-channel spectrum feature matrix; inputting the dual-channel spectrum feature matrix into a pre-trained shared encoder, extracting emotional feature representation of the individual, and obtaining an emotional state recognition result according to the emotional feature representation; according to the method, the depression emotion state of a new individual can be evaluated without participation of target domain data, the robustness and recognition precision of the model under cross-individual and low-labeling conditions are effectively improved, and the method has relatively high popularization and application potential.

The invention provides a multi-frequency biological resistance antibody component prediction method and device and a storage medium, and relates to the technical field of biomedical signal processing. According to the method, multi-frequency current testing is carried out on different physiological segments of a subject, an initial training data set is constructed in combination with basic information of the subject, the fine capturing ability of the multi-frequency bioelectrical impedance technology on human body physiological information is brought into full play, and individual features are fused to improve data dimensions and relevance; the quality of training data is guaranteed through data preprocessing, and a reliable foundation is laid for model training; and a component prediction model is constructed and optimized by a multi-task regression loss function, so that volume component prediction indexes can be efficiently output, and the prediction accuracy and stability of different indexes can be improved. According to the method, the problems of single data dimension and insufficient model precision in traditional volume component prediction are effectively solved, and a systematic and efficient scheme is provided for precise volume component prediction.

The invention relates to the technical field of biomedical signal processing and intelligent detection, and particularly discloses a PPG peak detection method based on a signal quality mask and dual-stage verification, and the method comprises the steps: obtaining an original PPG signal sampling sequence and a three-axis acceleration signal, and carrying out the data preprocessing of the original PPG signal sampling sequence; calculating a signal quality index SQI per second, and calculating a body motion variable coefficient CV; carrying out linear fusion on the SQI and CV to obtain a comprehensive quality score Score model; screening local candidate peaks according to a sliding window; backward verification and dynamic correction are carried out; updating a start bit of the sliding window according to the physiological rhythm, confirming a peak value and counting the peak value into a peak value list; setting a score threshold, and rejecting abnormal peaks in the peak list according to a Score model; according to the method, the candidate peak value is searched through the signal quality mask and the sliding window, and a two-stage verification mechanism of the final candidate peak value is determined through backward verification, so that the peak value detection accuracy is high, the motion interference resistance is high, and the real-time performance is good.

The invention provides a sleep physiological parameter real-time monitoring method based on flexible piezoelectric sensing, and relates to the technical field of biomedical signal processing.The method comprises the steps that a human body sleep biological mechanical signal is collected through a flexible piezoelectric sensor array, and after the signal is converted through a charge amplifier, phase-space reconstruction denoising and self-adaptive gain control are carried out; wherein the gain is dynamically adjusted based on the weight of the user, the rigidity coefficient of the mattress and the real-time signal amplitude; extracting heartbeat wave crest intervals, respiratory cycles and body movement frequency domain energy characteristics from the preprocessed signals; inputting the features into a cascaded convolutional neural network and a long-short-term memory network model, and synchronously outputting the heart rate, the respiration rate and the heart rate variability; in combination with the historical health data of the user and the cardiovascular health mapping model, sleep quality indexes and disease risk scores are generated through graph convolutional network analysis, and the monitoring frequency is dynamically adjusted and an apnea alarm is triggered.

The invention discloses a multi-dimensional night urination behavior monitoring system based on rhythm features, and relates to the technical field of biomedical signal processing, and the system specifically comprises a data acquisition module, a rhythm feature extraction module, a urine volume dynamics modeling module, a clustering monitoring module and a real-time feedback module; collecting individual data in real time through IoT equipment and preprocessing the individual data; calculating a nocturnal urination frequency index, a nocturnal urination time concentration ratio, a urination interval rhythm variation coefficient and a deviation index based on the individual data, and performing rhythm feature extraction; performing quadratic polynomial least square fitting on the night accumulated urine volume, and calculating a urine volume acceleration index by using an obtained second derivative to quantify a urine volume generation trend; the method comprises the following steps: constructing and preprocessing a night urine multi-dimensional digital phenotypic vector matrix, fitting a Gaussian mixture model based on an expectation maximization algorithm of a Bayesian information criterion, automatically mapping individuals into four types of subtypes according to cluster center features, and outputting individual subtype labels and confidence coefficients; and obtaining a comprehensive risk score through normalized risk assessment.

The invention discloses an emotion classification method and system for heart electromagnetic signals of heart rate variability based on topology analysis, and relates to the technical field of biomedical signal processing and emotion calculation, and the emotion classification method comprises the following steps: preprocessing magnetocardiogram signals or electrocardiogram signals; positioning R waves, extracting an RR interval sequence, and extracting statistics of topological feature points of the persistent graph as topological features; and performing fusion through the time sequence branch, the topological branch and the fusion layer to generate a fusion vector, and classifying the emotional state based on the fusion vector. According to the method, the signal quality and the anti-interference capability are remarkably improved, topological data analysis (TDA) is innovatively introduced, quantitative characterization of a heart rate variability nonlinear dynamic structure is achieved, and the capturing capability of the model to a key physiological mode is enhanced; the problem of overfitting under small samples is effectively relieved, the accuracy, robustness and cross-individual generalization ability of emotion classification are remarkably improved, and a new method is provided for emotion calculation.

The invention relates to the technical field of biomedical signal data processing, and particularly discloses a data processing method and system for a blood pressure electrical stimulator, and the method comprises the steps: synchronously collecting an electrocardiosignal, a blood pressure waveform signal and an electrical stimulation parameter signal of a testee at a moment through multiple channels; extracting an electrical stimulation switch moment sequence; constructing a standardized individual response data vector; establishing an individualized physiological response curve model by adopting a generalized linear regression method; and next stimulation parameters are adjusted through a feedback adjustment mechanism. In the prior art, an electrical stimulation method which depends on a fixed rule or empirical parameter setting, especially in an application scene in which physiological response differences exist among individuals, high-resolution physiological response data modeling cannot be realized. Due to the fact that data segmentation, low-sample modeling and predictive regulation and control can be carried out based on event driving, an individualized stimulation and response model is constructed, and the flexibility and safety of control over the blood pressure electrical stimulator are improved.

The invention provides an epilepsy prediction system based on multivariate weighted joint recursion and a graph attention network. The epilepsy prediction system can be applied to the technical field of biomedical signal processing and artificial intelligence. The system comprises a brain function imaging module which is configured to obtain electroencephalogram signal data of a target object under the condition that the target object is authorized; the processor comprises a multivariate weighted joint recursion processing unit which is configured to obtain a phase-space trajectory vector of each channel of the electroencephalogram signal data according to the electroencephalogram signal data, construct a recursion plot according to the phase-space trajectory vector of each channel, and calculate the phase-space trajectory vector of each channel based on the recursion plots of any two channels. Obtaining a channel correlation coefficient between any two channels, and according to the channel correlation coefficient between any two channels, constructing a weighted adjacency matrix representing the brain function of the target object; and the image attention network processing unit is configured to perform spatial-temporal feature processing according to the weighted adjacency matrix and the electroencephalogram signal data and output an epileptic seizure prediction result.

The present invention provides a device (100) and method for measuring and / or monitoring auscultation and other biomedical signals of a body part of a subject, the device (100) comprising: a plurality of sonographic sensors sensitive to surface vibrations-PG sensors (110) configured to measure acoustic activity of the body part, and providing a PG signal (115) indicative of acoustic activity of the body part, the PG sensor (110) being adapted to be arranged on the body part to cover different independent auscultation areas of the body part. The device further comprises at least one other biomedical sensor-BIO sensor (120) configured to independently measure a respective activity of the body part according to the sensor type and to provide at least one BIO signal (125) indicative of the respective activity of the body part. The device further comprises a signal processing unit (160) configured to acquire a signal (115, 125) from each sensor (110, 120) and to collectively perform a signal source specific analysis by combining information of the PG signal (115) with corresponding arrangement information of the PG sensor (110) on the body part and information of the BIO signal (125).

The invention relates to the technical field of biomedical signal processing and medical electronics, in particular to a blood pressure measuring method for non-invasive exercise blood pressure monitoring. The method aims to solve the technical problem of low blood pressure measurement accuracy caused by broadband noise interference and failure of a fixed parameter model in a motion state in the prior art. According to the technical scheme, the method comprises the steps that a cardiac cycle reference signal and a Korotkoff sound signal are synchronously collected, a sampling window is calculated in a self-adaptive mode by utilizing a cycle triggering feature point gating mechanism and combining a heart rate, and time domain precise interception of the signals is achieved; carrying out root mean square calculation and Savitzky-Golay filtering on the intercepted signal, and keeping waveform characteristics while denoising; constructing a multi-dimensional feature regression model containing a peak value, skewness, kurtosis and dynamic template similarity so as to accurately judge systolic pressure and diastolic pressure; an automatic recharging mechanism is triggered by monitoring signal quality, and strong interference is physically avoided. The method has the advantages of being high in anti-noise capacity, high in feature extraction robustness and accurate and reliable in measurement result.

The invention discloses a motor imagery electroencephalogram signal classification method and device, a terminal and a storage medium, and relates to the field of biomedical signal processing.The method comprises the steps that electroencephalogram signals based on user motor imagery are obtained and preprocessed, and electroencephalogram features are determined; performing multi-scale spatio-temporal feature extraction and space and channel decoupling reconstruction on the electroencephalogram features, and determining target spatio-temporal enhancement features; and classifying the target space-time enhancement features through a classification output layer, and determining a classification result. Due to the fact that space and channel decoupling reconstruction is carried out on the features, redundant correlation of the cross-electrode electroencephalogram signals is systematically eliminated, and the problems that in the prior art, space and channel information are jointly processed, information redundancy is caused, and calculation burden is increased can be effectively solved.

The application provides an epilepsy prediction system based on multi-element weighted joint recursion and graph attention network, which can be applied to the fields of biomedical signal processing and artificial intelligence technology. The system comprises a brain function imaging module configured to acquire electroencephalogram signal data of a target object under authorization of the target object; a processor comprising a multi-element weighted joint recursion processing unit configured to obtain a phase space trajectory vector of each channel of the electroencephalogram signal data according to the electroencephalogram signal data, construct a recursion graph according to the phase space trajectory vector of each channel, obtain a channel correlation coefficient between any two channels based on the recursion graph of each channel, and construct a weighted adjacency matrix representing the brain function of the target object according to the channel correlation coefficient between any two channels; and a graph attention network processing unit configured to perform spatiotemporal feature processing according to the weighted adjacency matrix and the electroencephalogram signal data and output an epilepsy seizure prediction result.

This invention provides a method for detecting abnormal signals in electroencephalograms (EEGs) based on step-by-step identification and multi-agent decision-making, belonging to the field of biomedical signal processing technology. The method includes: acquiring EEG signal data from epilepsy patients; parsing the data according to a preset channel order and segmenting it into multiple signal segments of equal length; extracting multi-dimensional features from each signal segment; inputting the temporal, frequency, and inter-channel synchronization features of each signal segment into an isolated forest model, and filtering for at least one potential abnormal segment based on an abnormality ratio threshold; performing multi-agent integrated decision-making on each potential abnormal segment based on its depth scattering, wavelet transform, temporal, and frequency features to obtain a comprehensive abnormality score corresponding to each potential abnormal segment; and determining the abnormal signal in the EEG signal data based on the comprehensive abnormality score corresponding to each potential abnormal segment. This invention significantly reduces the false alarm rate of abnormal signal detection in EEGs.

The application discloses a motor imagery electroencephalogram signal denoising method and device, medium and product, relates to the technical field of deep learning and biomedical signal processing, and the method comprises the following steps: acquiring a target electroencephalogram signal containing artifacts; inputting the target electroencephalogram signal containing artifacts into a trained electroencephalogram denoising model to obtain a final denoised electroencephalogram signal; wherein the electroencephalogram denoising model comprises an electroencephalogram denoising branch, an artifact prediction branch and an artifact representation interaction attention fusion reconstruction module; the electroencephalogram denoising branch comprises a multi-scale self-adaptive enhancement module, a frequency domain dynamic enhancement module and a feature extraction module. The application solves the problems of traditional electroencephalogram denoising methods in signal fidelity, spectral fidelity, spatial structure preservation and generalization ability.

The invention relates to the technical field of biomedical signal acquisition, and discloses a portable electroencephalogram acquisition system, which comprises an inertia sensing unit, a signal processing unit, a signal processing unit, a signal processing unit and a signal processing unit, and is characterized in that the inertia sensing unit is used for monitoring head movement data and outputting interruption when detecting a movement event; the main control module responds to the interruption to calculate an instantaneous motion intensity index, triggers an impedance monitoring mechanism when the index exceeds a threshold value, and controls the analog front-end module to inject alternating current test current; and the main control module calculates an equivalent impedance module value according to the collected mixed signal, and judges whether the interference type is friction interference or contact failure in combination with a motion index. And the master control module sends an instruction to control the analog front end to execute hardware strategies such as right leg driving gain adjustment or input channel transient short-circuit protection. According to the invention, through a multi-dimensional physical quantity joint judgment and hierarchical hardware anti-interference mechanism, source suppression of noise of different causes is realized, saturation of the amplifier is prevented, and the signal stability of the portable equipment in a motion state is improved.

The disclosure is directed at methods and systems using a metasurface to enhance radar sensing. The system of the disclosure may be seen as a system that provides high near-field sensing in radar systems for biomedical signal monitoring or sensing applications. In some embodiments, the disclosure is directed at a metasurface that combines low profile and high integration capability to accommodate mm-wave frequencies.

The invention belongs to the technical field of biomedical signal processing, and discloses a fatigue recognition model training method and system based on three-lead electroencephalogram. Existing fatigue recognition depends on multi-lead equipment, the cost is high, and dynamic time-frequency evolution information of fatigue cannot be captured during feature extraction. According to the technical scheme, the method comprises the steps that three-lead electroencephalogram signals are preprocessed and subjected to fatigue degree marking, then, time sequence features, frequency domain features and time-frequency domain features are extracted from electroencephalogram data, then, the time sequence features are utilized to train a lightweight LSTM model, the frequency domain features are utilized to train an XGBoost model, and the time-frequency domain features are utilized to train a CNN-LSTM fusion model. According to the method, through mutual cooperation of the three-lead electroencephalogram data and the reasoning model, accurate recognition of the fatigue degree of the three-lead portable equipment is achieved, and a feasible path is provided for large-scale and normalized fatigue risk monitoring of high-risk workers.

The application provides a training method and system of an emotion classification model for multi-modal physiological signals, relates to the technical field of cross between artificial intelligence and biomedical signal processing, and the method comprises the following steps: collecting multi-modal physiological signals of a plurality of sample users; processing the multi-modal physiological signals of the plurality of sample users to obtain multi-modal physiological vectors of the plurality of sample users; using the multi-modal physiological vectors of the plurality of sample users to perform autoregressive pre-training on a large language model; using the multi-modal physiological vectors of the plurality of sample users and text vectors corresponding to emotion classification prompt texts to fine-tune the large language model subjected to the autoregressive pre-training, and obtaining an emotion classification model. In the process of emotion prediction, the problems of multi-modal signal mode loss, inconsistent sampling rates and non-uniform channel numbers can be effectively overcome, and the accuracy of emotion prediction is improved.