The Analysis Of Surface Electromyographic Signal Based On Wavelet Transform And Multifractal Analysis

Surface electromyographic (SEMG) signals can be monitored noninvasively by using electrodes on the skin surface. They are the summation of all motor unit action potentials (MUAP) within the pick-up area of the electrodes, so they provide information of the neuromuscular activities of the examined muscle. In addition, these signals have been widely applied to clinical diagnosis, sports medicine, ergonomics, rehabilitation medicine, neurophysiology and electrophysiology. Aimed at control system for powered prostheses and muscle fatigue assessment, this dissertation deeply investigated SEMG signals using wavelet transform and nonlinear dynamic method. The main and creative work is as follows:In order to improve the classification accuracy of SEMG signals in control system for powered prostheses, we tried to use local discriminant bases method to classify SEMG signals and the main idea of this method is to employ wavelet packet decomposition coefficients with maximum class separability as the feature vectors. However, considering that the energy of the SEMG signals varies with different subjects and different movements, this dissertation presented the optimal wavelet packet method based on discriminant measures and relative energy representation of wavelet packet and applied this method to SEMG signals classification. Experimental results showed that this approach achieved higher accuracy than fixed scale wavelet packet method. In this method, the discriminant measures which we used are only some simple distance measure function and they can not give the optimal features for pattern recognition problems. So we improved the above method and presented the optimal wavelet packet method based on Davies-Bouldin (DB) index and relative energy representation of wavelet packet, which was then applied to SEMG signals classification. Compared with other existing methods, this one had significant improvement in classification accuracy.In order to reduce the computation time of classifying SEMG signals in control system for powered prostheses, this dissertation proposed an efficient method to SEMG signals classification based on the discrete harmonic wavelet packet transform (DHWPT). Firstly, the relative energy of the signal in each frequency band calculated after the signal had been decomposed by the DHWPT was used as features of a SEMG signal. Then, the feature selection method based on the genetic algorithm and the neural network classifier were employed to provide the best discriminating features of different categories of movement. In the end, a neural network classifier used these selected features to validate the classification performance of the presented method. Compared with other methods of SEMG signals classification, the DHWPT method possessed higher classification accuracy than the time domain method and saved more computational time than the discrete ordinary wavelet packet transform method. So the DHWPT method was an efficient approach to classifying SEMG signals.The systematic study of muscle fatigue assessment can provide sight into the physiology of the muscle under investigation as well as the mechanisms of fatigue. This dissertation used multifractal method to analyze SEMG signals in the course of muscle fatigue. At present, we have found that the SEMG signals characterized multifractality during static contractions and the area of the multifractal spectrum of the SEMG signals significantly increased during muscle fatigue. Therefore the area could be used as an indicator of assessing muscle fatigue during static contractions. Compared with the MDF which was the most popular indicator for assessing muscle fatigue, the spectrum area presented here showed higher sensitivity. So the singularity spectrum area was considered to be a more effective indicator than the MDF while estimating muscle fatigue during static contractions. During dynamic contractions, the SEMG signals still characterized multifractality. At the same time, we found that the area of the multifractal spectrum of the SEMG signals also increased during muscle fatigue. Hence the area could also be used as an indicator of assessing muscle fatigue during dynamic contractions. However, the slope of the singularity spectrum area during dynamic contractions is smaller than that during static contractions. We thought that the difference between static contractions and dynamic contractions is caused by the blood flow in the contracting muscle.