Frameworks for u-Health Bio Signal Controller

Summary

Control technologies based on biosignal manipulate devices such as computer and wheelchair. ElectroMyoGram(EMG), ElectroOculo- Gram(EOG), ElectroEncephaloGram(EEG) are typical important bio-signals for u-health care. In this paper, we approach the EMG signals from electrodes placed on the forearm and recognizes the four kinds of motion. We also develops the prototype of a u-health device controller that controls hardware devices using EMG signal. To analyze EMG with properties of non-stationary signal, time-frequency features are extracted by wavelet packet transform. For dimensionality reduction and nonlinear mapping of the features.We proposes a feature projection method composed of PCA and SVM. The dimensionality reduction simplifies the structure of the classifier, and reduces processing time for the pattern recognition. The nonlinear mapping using SVM transforms the PCA-reduced features to a new feature space with a highly class separatability. SVM is a pattern classifier to recognize various motions.We finally show the experimental results using the proposed method enhances the accuracy of pattern recognition. As a results, The proposed systems make the possible to control movements of u-healthcare signal device based on classified patterns.