| Electromyography (EMG) has attracted attention in both academic research and medical diagnosis, due to its wide application in neurology, ergonomics, and physiology. There are intramuscularly recorded needle EMG signals and surface recorded EMG signals. In this work, we focus on the surface EMG (SEMG) signal.;Onset detection for SEMG signal is important due to its various applications. In this work, we reviewed four popular onset detection approaches: Hodges, Bonato, Lidierth, and Abbink; and evaluate their performance using real and synthetic SEMG signals. We analyzed the influence of high pass filters to the onset detection accuracy. We also propose a simple and generally applicable mechanism to determine an optimal threshold for various detection functions. Experimental results demonstrate that the proposed mechanisms (high pass filtering and a method for determining an optimal threshold) greatly improve the onset detection accuracy, which is comparable to manual onset detection. |
