Time-Varying Multicomponent Signal Modeling for Analysis of Surface EMG Data

We present a novel approach to surface EMG data characterization by using time-varying multicomponent signal modeling. An EMG signal is described as a set of stationary non-harmonically related sinusoids (signal components) whose time-varying bandwidth is modeled by polynomials. The polynomial coefficients, estimated from a set of linear equations, capture the relationship between the instantaneous frequency and amplitude for individual signal components. It is proposed that such a compact EMG signal modeling may be a good candidate for a number of applications in surface electromyography: compression, muscle activity detection and low-bias conduction velocity, to name a few.