Measurement of Walking Speed from EMG Signal using Kurtosis of Approximate Coefficients

摘要

In this paper, the speed of human walking has been assessed from EMG signals using Kurtosis of approximate coefficients-based wavelet transform. The signals have been captured from three muscles of lower extremity of normal individuals during level walking at different speeds. The collected signals have been denoised as the signals were heavily noised. The denoised EMG data have been decomposed at various DWT levels for feature extraction and subsequently Kurtosis values have been determined. The patterns of change of Kurtosis values with respect to walking speeds corresponding to individual DWT levels have been studied and compared. Results demonstrated many definite relations between walking speeds and Kurtosis values. Most regular patterns have been utilized for scaling. Finally, an algorithm has developed for assessment of unknown walking speed data and subsequently validated. EMG sensor-based different scheme for monitoring health parameters of disable persons like muscle strength, response, etc., are becoming familiar in biomedical image processing. The method proposed in this work is adding another effective means of monitoring walking speed using same type of sensors which incorporates both time-frequency domain analysis. Therefore, the method seems a little bit complex; however, it can be used in a compact and robust way with other existing schemes for monitoring health parameters. The work can be extended to inter-link the walking speed features with those parameters and can be utilized in closed loop prosthetic control having EMG sensory feedback.