Improved Prosthetic Control Based on Myoelectric Pattern Recognition via Wavelet-Based De-Noising

Abstract

Real-time inference of human motor volition has great potential for the intuitive control of robotic devices. Toward this end, myoelectric pattern recognition (MPR) has shown promise in the control of prosthetic limbs. Interfering noise and susceptibility to motion artifacts have hindered the use of MPR outside controlled environments, and thus represent an obstacle for clinical use. Advanced signal processing techniques have been previously proposed to improve the robustness of MPR systems. However, the investigation of such techniques have been limited to offline implementations with long time windows, which makes real-time use unattainable. In this work, we present a novel algorithm using discrete and stationary wavelet transforms for MPR that can be executed in real-time. Our wavelet-based de-noising algorithm outperformed conventional band-pass filtering (up to 100 Hz) and improved real-time MPR in the presence of motion artifacts, as measured by the motion test. Improved signal-to-noise ratio was found not to be crucial in offline MPR, as machine learning algorithms can integrate high but consistent noise as part of the signal. However, varying interference is expected to occur in real life where signal processing algorithms, as the one introduced in this paper, would potentially have a positive impact. Furthermore implementation of these algorithms in a prosthetic embedded system is required to validate their feasibility and usability during activities of the daily living.

Denoising of surface electromyogram based on complementary ensemble empirical mode decomposition and improved interval thresholding

“…Stationary wavelet transform (SWT)-based denoising has been proven to derive good sEMG signals. 30 Wavelet thresholding can be applied directly to EMD, but EMD direct thresholding (EMD-DT), whether hard or soft, can have catastrophic consequences. 24 IT is proposed to overcome the discontinuity of direct thresholding.…”

Effectiveness of Surface Electromyography in Pattern Classification for Upper Limb Amputees

“…The Daubechies db7 wavelet was selected based on related DWT usage with EMG [3] and comparison with the signal characteristics of the EMG datasets in use. For decomposition, 3 and 4 levels are common, cited as optimum [3] [12]. Four levels were selected, implying four detail subbands, and a corresponding 4th level approximation subband.…”

“…Using wavelet analysis, a signal is represented as a series of oscillatory functions of finite duration (wavelets) by decomposing it using a wavelet transform. It can therefore be expressed as a linear combination of these functions with wavelet coefficients giving a compact representation of the signal's energy [12]. Significant work has been done in recent decades, to elucidate ideal feature choice [14], [15], [16].…”

A Self-Selected Movement Classification Method for Forearm Via sEMG and Attitude Sensor

“…In order to achieve the accuracy of the control, it is necessary to improve the recognition accuracy of the movements. For some fixed movements, through long-term training, more than 99% of the accuracy can be achieved [4,5] . However, this method is mainly for the fixed movements, once the subjects are replaced, the accuracy will drop significantly.…”

“…Oluwarotimi et al [14] made use of a given Analysis Window and its Mean (ASM) to get features with accuracy of 92%. Julian et al [4] put to use wavelet transform to extract features. There are also many extraction methods, including: Standard Deviation (SD), Waveform Length (WL), Median Frequency (MDF), Slope Sign Change (SSC), Power Spectrum density (PSD), and so on.…”