Neural network-based classification of electromyographic (EMG) signal during dynamic muscle contraction

Abstract

A neural network-based decision making tool was developed for on-line classification of various neuromuscular diseases. Intramuscular electromyographic (EMG) signals were collected using microcomputer-based real-time data acquisition system during voluntary human muscle contraction. The data was collected from three different group of patients. A number of signal processing techniques such as: Autoregressive (AR) modeling, Short Time Fourier Transform (STFT), Wigner-Ville Distribution (WVD), and Chaos analysis methods were applied to these signals. A pair of signal processing features extracted from each of these methods was used to train a neural network using backpropagation algorithm. MATLAB based software was developed and interfaced with Aspirin/MIGRAINS, a neural network develop ment and training tool. A set of data from eight patients of three Merent patient groups (normal, neuropathy, and myopathy) was used for training and testing the neural network. The signal processing features h m six data of each different known patient group were used as inputfoutput to train the network. The rest two sets of the data of each patient group were used to check the validity of the classification made by the network. The results agree with other methods of classiikation and are promosing for on-line dassiikation and diagnosis of various neuromuscular diseases.

Application of the Chaos Theory in the Analysis of EMG on Patients with Facial Paralysis

“…Chaos is a complex motion which is always restricted in limited areas, extremely sensitive to initial values, long-term unpredictable, with track never repeated, fractal dimension and strange attractors. Amount of study demonstrated that Center Nervous System (CNS) generates chaotic firing patterns of action potentials; and a variety of physiological potential nerve signals, including EEG, ECG and EMG, etc., originated from CNS, have shown some degree of chaotic behavior, too [16][17][18][19][20][21][22][23][24][25][26][27][28]. Some nonlinear analysis methods are applied to the analysis of EMG.…”

“…Bodruzzaman et al extracted fractal dimension [18], Hurst's rescaledrange, Housdorff-Besicovich fractal dimension [19] and correlation dimension [20] respectively, to analyze sEMG acquired from the patients with neuromuscular diseases. The results indicated that chaotic features of sEMG are effective indicators for diseases diagnosis.…”