Objective: To evaluate an adaptive-filter method for use in analysis of periodic electromyography (EMG) signals in which the transfer function of the filter is matched to characteristics of the signal.
Animals: 15 adult horses without clinical signs of back pain.
Procedure: Electromyography signals of the left and right longissimus dorsi muscles, middle gluteal muscles, and triceps brachii muscle were recorded from horses walking on a treadmill, using bilaterally placed surface electrodes. A reflective marker was placed on the hoof of the left hind limb for simultaneous kinematic measurement of motion cycles. Absolute value of the measured EMG signal was convoluted by use of a filter signal equivalent to the length of 3 motion cycles. The signal-to-noise ratio (SNR) was calculated from the autocorrelation function and compared with the SNR of the unfiltered and the low-pass filtered signals.
Results: The signal-adapted filter significantly increased SNR (by 7.3 dB, compared with the low-pass filter, and by 11.1 dB, compared with the unfiltered EMG signal).
Conclusions and clinical relevance: The signal-adapted filter eliminates signal parts that are not correlated to periodic motion. The method reported here improves the applicability of periodic EMG signals as a clinical tool.