Some Aspects of Nonstationary Myoelectric Signal Processing

Nonstationary myoelectric signal processing is considered. A midpoint moving average estimator for the signal variance is developed. The performance of the estimator is analyzed and optimum window size criteria are derived. Comparison of the theoretical performance to experimental performance shows good agreement. The resultant estimation error is only moderately (2-3 percent) degraded from the equivalent stationary case. Finally, the developed estimator is compared to a typical squarer followed by a low-pass filter myoelectric processor. Provided that the low-pass filter has a linear phase characteristic in the passband, the two systems have essentially the same error performance.