Research And Application Of Scale Division Method In Lemple-Ziv Complexity

Recognition of human emotion plays an increasingly important role in the aspects of the daily application. The resulting EEG becomes one of the primary means in human emotion study.Complexity is an important method to research EEG feature extraction. However, the process of the binary of the traditional complexity some limitations that the staging area is previously fixed, which cannot depicts the relationship between adjacent points and details.In order to overcome these limitations, we proposes improved adaptive Lempel-Ziv complexity algorithm and explore the variation under different emotional states and different electrode complexity, and use SVM to classify the feature, and verify the quality and effectiveness of the feature extraction. Specific work as follows:(1) According to the characteristics of EEG emotion which is to the study, we select for nonlinear dynamic extraction mode, and discuss the feasibility of the algorithm complexity of EEG feature extraction. Using the traditional algorithm and multi-scale complexity algorithm to feature extraction of the motor imagery data, and use support vector machine(SVM)to evaluate the results. The results show that the complexity of the algorithm based on nonlinear dynamics can effectively for the motor imagery EEG feature extraction, Lempel-Ziv complexity can be used as an effective feature extraction method for EEG data analysis.(2) Combined with the characteristics of EEG emotion data, the analysis of the traditional complexity algorithm and multi-scale complex algorithm, on their foundations to adapt Lempel-Ziv complexity algorithm which is improved, we use the adaptive Lempel-Ziv complexity algorithm for the motor imagery EEG feature extraction, comparison of classification accuracy rate of motion picture feature extraction which is under different complexity algorithm.(3) At the same time we use the adaptive Lempel-Ziv algorithm to extract the feature of EEG emotion data. According to the subjects and music videos, and the results were statistically analyzed and compared with the traditional Lempel-Ziv algorithm and Lempel-Ziv complexity at multi-scale, we get the complex changes of different electrode in different emotional states, and then draw the activation status in different brain regions under different emotion.