Main Article Content

Abstract

Electromyograph signal (EMG) is a non-stationary biomedical signal, making it difficult to
determine the pattern. The method normally used for signal analysis is Fast Fourier Transform (FFT), but it has some drawbacks because it requires stable signals. To answer this deficiency wavelet transformation is used, especially discrete wavelet transforms that can analyze the signal in both the realm of time and frequency.
The method to be used in this research is wavelet transformation for signal analysis with
decomposition up to level 7 using wavelet symlet 8. This feature extraction result is used as input of artificial neural network (ANN) type of propagation backward with architecture of 8 input layer, 5 hidden layer and 3 layers of output.
ANN Turnback is able to recognize 3 types of EMG signals namely Normal, Myopathy and
Neuropathy. Based on the feature extraction of EMG signal decomposition energy characteristics. Network architecture with 8 input layers. 5 hidden layers and 3 output layers Proven best in the introduction of EMG signals. The highest success rate is the introduction of EMop Myopathy signal pattern reaching 94%, so the network architecture is proposed to regenerate the EMG signal.

Article Details