Research On Emotion Recognition And Sleep Staging Based On Physiological Signals

In today's society,with the continuous improvement of the pace of life and the increasingly fierce competition in work,most people are under the pressure of both mental and physical.Scientific research shows that people who are under high stress for a long time are prone to negative emotions such as disgusting,depressed,and tempered.Moreover,such people are more likely to develop physical or mental diseases such as insomnia,depression,etc.,which may cause cardiovascular diseases,human metabolism and mental disorders.Based on the traditional psychological self-rating scale method,there is a certain subjectivity,and the effective evaluation and recognition of emotional state through physiological signals can help the psychologist to understand the emotional state of each individual.Based on different emotional states,the corresponding psychological guidance and treatment are carried out to help specific depressed patients adjust their mood and restore their health.Sleep is one of the most important physiological activities of human beings,and its quality directly or indirectly affects the health of the human body.Through sleep monitoring of physiological signals,each person's sleep stage,sleep structure and sleep quality can be analyzed.Accurate sleep staging monitoring can help doctors analyze and diagnose sleep,and find out the rules for the deeper etiology of sleep diseases.In this paper,the evaluation of emotional state based on physiological signals and sleep staging is the research goal.The physiological signal processing and data analysis are the core,and the single-channel blood oxygen signal PPG and five-channel EEG are compared.Accurate emotion recognition and more accurate sleep staging results based on single-channel ECG signals.The main work of this paper is:Part ?: The internationally recognized public dataset DEAP sentiment dataset used as a data source for emotion recognition.In order to better integrate the existing wearable technology application features,use fewer signal channels to achieve better classification results.Finally,the PPG signal and the EEG signal in the data set are selected.The signal processing and feature extraction of PPG signal and EEG signal in data set are respectively carried out,and the logistic regression algorithm and Adaboost algorithm are used for classification modeling,and the system parameters of the model are adjusted to obtain better classification accuracy of emotion recognition.Part ?: sleep staging uses the internationally recognized library MIT-BIH POLYSOMNOGRAPHIC DATABASE as the data source,and selects the single-channel ECG signal in the data set as the signal source,performs signal processing on the ECG signal,extracts the eigenvalue;and passes the peak of the ECG signal.The interpolation signal is used to fit the EDR signal to obtain the characteristic value of EDR.Cardiopulmonary coupling analysis of ECG and EDR signals was performed to obtain coupling characteristics,and the complex features of ECG signals were further extracted to improve the final sleep staging accuracy.The random forest algorithm is used to screen the features,and the optimal feature subsets are obtained.Then the SVM algorithm is used to model the classification system,and finally the two classifications(WAKE-REM/NREM)and the three classifications(WAKE-REM-NREM)are respectively obtained.It can reach 90.48%,88.10% of healthy people and 83.74%,78.54% of patients with sleep disorders,which is the best result of sleep staging that is known to utilize single-lead ECG signals and use the same data set.