Emotion Recognition Based On Physiological Signals And Its Application In Hand Function Rehabilitation Training

Nowadays,with the development of artificial intelligence technology,the application of human-computer interactive rehabilitation robots in the medical field is more and more extensive.Relevant studies have shown that in the rehabilitation treatment for stroke patients,patients are vulnerable to psychological and emotional interference,which affects the rehabilitation training of patients by rehabilitation robots.Therefore,on the psychological level,it is very urgent to carry out research on patients' active rehabilitation training.Based on the analysis of physiological signals,this paper carries out the research on emotion recognition and rehabilitation training of hand function.Taking excitement,irritability and relaxation as the target emotions,using human EMG and ECG physiological signals,an emotion recognition system based on physiological signals was studied and applied to hand function rehabilitation training.Firstly,the features of EMG and ECG were selected by the improved Relief F algorithm.Then,the BP neural network optimized based on Ant-lion algorithm(ALO)is used to realize emotion recognition of physiological signals.Finally,an online emotion recognition system based on STM32 is designed and implemented and applied to the hand rehabilitation training robot.The specific work is as follows:(1)Human physiological signal preprocessing and improved Relief F algorithm feature selection.Firstly,the EMG and ECG signals are preprocessed by noise reduction,and the features are extracted in time domain and frequency domain;because the collected physiological signals are easy to extract abnormal signal features due to body movement,and the traditional Relief F algorithm selects features according to the shortest distance of the sample,it can not represent the characteristics of all features.To solve the above problems,this paper adds abnormal feature judgment on the basis of traditional Relief F algorithm,and then selects feature samples according to the distance proportion of feature samples to calculate feature weights.Finally,according to the weight threshold,the features with higher weight are analyzed for crosscorrelation,and the features with higher correlation are removed to generate a new feature subset.This method can effectively eliminate redundant features and interference features,and improve the effect of emotion recognition.(2)Design and implementation of human physiological signal emotion recognition model based on BP neural network optimized by ALO.Because the traditional BP neural network algorithm is easy to fall into local minimum in the determination of network weights and thresholds.To solve this problem,this paper uses ALO to optimize the initial weights and thresholds of BP neural network.The ALO-BP neural network is applied to the emotion recognition in this paper,and the neural network structure of the emotion recognition system is systematically elaborated.Through the comparative experiment,the performance of the network model is analyzed,and it is found that the recognition rate of the ALO-BP neural network is improved.(3)Design and implementation of emotion recognition system and experimental analysis of its application in hand rehabilitation robot.In this paper,STM32F103RET6 is used as the main control unit to design and implement the hardware and software of the emotion recognition system based on physiological signals,which is applied to the hand function rehabilitation robot.Through the physiological signal acquisition module of the emotion recognition system,the EMG and ECG physiological signals are collected and the features are extracted.The features are transmitted to the ALO-BP neural network on the PC side for optimization training.The optimized neural network weights and thresholds are written into the emotion recognition system to realize emotion recognition.The emotion signal was regraded as complementary control signal,which could be used to control the robot.This system could achieve consistent control between robot and patient and change the training strategy simultaneously.