Quantifying improvements in motor control is predicated on the accurate identification of the onset of surface electromyograpic (sEMG) activity. Applying methods from wavelet theory developed in the past decade to digitized signals, a robust algorithm has been designed for use with sEMG collected during reaching tasks executed with the less-affected arm of stroke patients. The method applied both Discretized Continuous Wavelet Transforms (CWT) and Discrete Wavelet Transforms (DWT) for event detection and no-lag filtering, respectively. Input parameters were extracted from the assessed signals.

The onset times found in the sEMG signals using the wavelet method were compared with physiological instants of motion onset, determined from video data. Robustness was evaluated by considering the response in onset time with variations of input parameter values.

The wavelet method found physiologically relevant onset times in all signals, averaging 147 ms prior to motion onset, compared to predicted onset latencies of 90-110 ins. Latency exhibited slight dependence on subject, but no other variables.