Bimodal Anxiety State Assessment Based on Electromyography and Electroencephalogram

摘要

Anxiety is a common emotional state of patients in rehabilitation training, which will affect rehabilitation training. In the current research, there are problems such as single use of electromyography signal to quantify anxiety and low accuracy. A method of bimodal feature fusion of electromyography (EMG) and electroencephalography (EEG) is proposed to realize anxiety assessment. First, the anxiety induced by 16 healthy subjects was recorded by EMG and EEG, comprehensive features were extracted from the signals. Then the pattern recognition model fused the features of EMG and EEG, based on theories of genetic algorithm, particle swarm optimization and support vector machine, which was designed to anxiety assessment by adjusting the feature fusion coefficient. According to the anxiety grading standard, three anxiety states were selected as severe, moderate and mild. Using only EEG data and EMG data as the sample sets, the corresponding recognition accuracy rates are 71.54% and 67.98%. However, using EEG and EMG feature fusion data as the sample sets, the average recognition accuracy rate was 81.49%, which increased by 9.95%, 14.01% compare to using EEG or EMG data individually. In short, the method proposed in this study improves the accuracy of anxiety state recognition and helps to better study patient anxiety.