Processing of Natural Signals like EMG for Person Identification using NUFB-GMM
Suresh M, P G Krishnamohan, Mallikarjun S Holi
Abstract
Physiological signals like Electrocardiogram(ECG) and Electroencephalogram(EEG), including deoxyribonucleic acid(DNA) are person specific and distinct for different persons. The motor unit firing pattern, motor unit recruitment order and characteristics of muscle changing from person to person, and therefore Electromyogram (EMG) can be used for person identification. EMG records obtained from a single channel data acquisition system are used to develop person identification system. Non-uniform filter bank (NUFB) technique used to extract features from EMG signal. The Gaussian Mixture Model (GMM) is used to generate person models from NUFB features. The EMG data of 100 healthy persons is recorded in different three sessions. So person identification is proposed using EMG signal and gives better result in performance. The performance of this person identification system for change in total number of persons substantially remains stable from Gaussians size of 64 onwards.
Keyword
Biometrics, Electromyography (EMG), Feature extraction, Gaussian mixture model (GMM), Non-uniform filter bank (NUFB).