Improvement in the algorithm for accurate detection of muscle activity from EMG

This paper presents the novel algorithm for accurate detection of muscle activity from EMG. The tibialis anterior muscle supports the human movement when the exercise with the tipping toes up method is performed. Ten healthy volunteers who are between 18 - 25 years old participate in determining the on-off EMG signal. The system consists of an amplifier with gain 981, CMRR at 91 dB, and a band-pass filter at the cutoff frequency of 50-500 Hz. Sampling frequency was 1000 Hz. Total EMG signal are 20 data sets. A window size was set at 2500 millisecond for each data set. The performance of the proposed algorithm, which is based on forward and backward standard deviation (FBSD), is compared with the conventional standard deviation (CSD) method using a mean absolute error value (MAE). Result shows that MAE values of the FBSD are better than those from the CSD method. That is, while the average MAEs of CSD are 2.85 and 155.9 for the on and off times of EMG, the average MAEs from FBSD are only 1.4 and 1.9, respectively.