Baseline adaptive wavelet thresholding technique for sEMG denoising
journal contributionposted on 21.05.2015 by Luca Bartolomeo, Massimiliano Zecca, Salvatore Sessa, Zhuohua Lin, Y. Mukaeda, Hiroyuki Ishii, Atsuo Takanishi
Any type of content formally published in an academic journal, usually following a peer-review process.
The surface Electromyography (sEMG) signal is affected by different sources of noises: current technology is considerably robust to the interferences of the power line or the cable motion artifacts, but still there are many limitations with the baseline and the movement artifact noise. In particular, these sources have frequency spectra that include also the low‐frequency components of the sEMG frequency spectrum; therefore, a standard all‐bandwidth filtering could alter important information. The Wavelet denoising method has been demonstrated to be a powerful solution in processing white Gaussian noise in biological signals. In this paper we introduce a new technique for the denoising of the sEMG signal: by using the baseline of the signal before the task, we estimate the thresholds to apply to the Wavelet thresholding procedure. The experiments have been performed on ten healthy subjects, by placing the electrodes on the Extensor Carpi Ulnaris and Triceps Brachii on right upper and lower arms, and performing a flexion and extension of the right wrist. An Inertial Measurement Unit, developed in our group, has been used to recognize the movements of the hands to segment the exercise and the pre‐task baseline. Finally, we show better performances of the proposed method in term of noise cancellation and distortion of the signal, quantified by a new suggested indicator of denoising quality, compared to the standard Donoho technique.
This research has been supported by the G-COE Global Robot Academia Program in Waseda University, Japan and partially by a Grant by STMicroelectronics. This research has been conducted at Humanoid Robotics Institute, in collaboration with the G-COE Global Robot Academia. The authors would like to express their gratitude to Okino Industries LTD, Japan ROBOTECH LTD, SolidWorks Corp, Dyden, for their support to the research.
- Mechanical, Electrical and Manufacturing Engineering