High-precision EMG signal decomposition with modified Viterbi algorithm

A novel high-precision decomposition algorithm for EMG (electromyography) signals based on the Viterbi algorithm is described. Under certain assumptions this algorithm yields the least squares solution in the time domain. It is shown that the EMG signal source can be modeled as a (M+1)-ary signaling system with intersymbol interference. Direct application of the Viterbi algorithm would lead to a decoder with ≈10²⁰ states, which cannot be implemented. It is shown that according to the special nature of the EMG signal essential simplifications can be done. The resulting algorithm requires only on the order of 10⁴ states. The test of the implemented algorithm shows that, in the mean, over 90% correctly detected innervation time points and only about 5% additional time points can be expected