Parametric sEMG Muscle Activity Detection Based on MAV and Sample Entropy

The electromyographic have become popular in many research areas, from human machine interface to biomechanics. Identifying which parts of the signal are muscle activity is a common problem faced by many applications that deals with electromyographic processing. Different kinds of solutions were proposed for this necessity, each of them presenting some advantages and disadvantages. This paper shows an alternative technique, which is window based using mean absolute value and sample entropy, both low consuming features extracted from the raw or pre-filtered EMG signal. Although this method requires some statistical parameters, it doesn't rely on threshold neither on frequency domain transformation. Tests under different circumstances were conducted, showed good muscle activity separation from baseline for real EMG signals, which collected from biceps of both men and women.