Improving the Robustness of Myoelectric Pattern Recognition for Upper Limb Prostheses by Covariate Shift Adaptation

Abstract

Fundamental changes over time of surface EMG signal characteristics are a challenge for myocontrol algorithms controlling prosthetic devices. These changes are generally caused by electrode shifts after donning and doffing, sweating, additional weight or varying arm positions, which results in a change of the signal distribution-a scenario often referred to as covariate shift. A substantial decrease in classification accuracy due to these factors hinders the possibility to directly translate EMG signals into accurate myoelectric control patterns outside laboratory conditions. To overcome this limitation, we propose the use of supervised adaptation methods. The approach is based on adapting a trained classifier using a small calibration set only, which incorporates the relevant aspects of the nonstationarities, but requires only less than 1 min of data recording. The method was tested first through an offline analysis on signals acquired across 5 days from seven able-bodied individuals and four amputees. Moreover, we also conducted a three day online experiment on eight able-bodied individuals and one amputee, assessing user performance and user-ratings of the controllability. Across different testing days, both offline and online performance improved significantly when shrinking the training model parameters by a given estimator towards the calibration set parameters. In the offline data analysis, the classification accuracy remained above 92% over five days with the proposed approach, whereas it decreased to 75% without adaptation. Similarly, in the online study, with the proposed approach the performance increased by 25% compared to a test without adaptation. These results indicate that the proposed methodology can contribute to improve robustness of myoelectric pattern recognition methods in daily life applications.

Transfer Learning for Rapid Re-calibration of a Myoelectric Prosthesis After Electrode Shift

“…Fortunately, our proposed transfer learning scheme significantly improves the performance again to near-perfect accuracy, outperforming the previously reported results by [12]. These results hold even for simultaneous movements in multiple degrees of freedom.…”

“…Muscle activity was recorded at 1000Hz sampling rate with an eight channel Ottobock Healthcare electrode array (13E200) attached around the forearm. To simulate doffing and donning of a prosthesis, the electrodes were shifted transversally by 8mm, as in [12], and all movements were recorded once more. Transversal shift was applied as it has been shown to be more challenging than longitudinal shift [6].…”

“…However, machine learning algorithms require signals corresponding to the same intended motion to be constant over time, while in practice the myoelectric signal is vulnerable to disturbances, such as electrode shifts, posture changes, sweat, fatigue, etc. [5,12]. As of today, these problems with respect to robustness are too severe to apply machine-learning algorithms as part of a commercial product [2].…”

“…However, a method that achieves robust control using only a moderate amount of surface EMG electrodes is still lacking [2]. Therefore, recent research has suggested an alternative approach, namely developing a re-calibration scheme that permits users to regain the high accuracy obtained under lab conditions, using only minimal training information (less than 1 minute of recording time) [12]. While promising, the current state-of-the-art is not yet able to fully regain the recognition accuracy obtained under lab conditions and does not yet allow simultaneous movements.…”

ML2Motif—Reliable extraction of discriminative sequence motifs from learning machines

“…Generally, it is not restricted by a specific form of application and/or learning machine. Hence, it could be easily applied to other types of applications such as explanation of most expressive electrode-combination in hand movement recognition with EMG signals [31], change point/anomaly detections in time series for fault detections in wind turbines [32, 33], explanation of important pixel patches in computer vision [6], quantum chemistry [34], and extraction of latent brain states [35]. However, there are two main shortcomings: first, it does not take any non-linear correlations of features into account and second, the number of samples depends on the complexity of the problem.…”