Adaptive Pattern Recognition to Ensure Clinical Viability over Time

Pattern Recognition is a useful tool for deciphering movement intent from myoelectric signals. In order to be clinically viable over time, recognition paradigms must be capable of adapting with the user. Most existing paradigms are static, although two forms of adaptation have received limited attention: Supervised adaptation achieves high accuracy, since the intended class is known, but at the cost of repeated cumbersome training sessions. Unsupervised adaptation attempts to achieve high accuracy without explicitly being told the intended class, thus achieving adaptation that is invisible to the user at the cost of reduced accuracy. This paper reports a novel adaptive experiment on eight subjects that allowed a post-hoc comparison of four supervised and three unsupervised adaptation paradigms. All supervised adaptation paradigms reduced error over time by at least 23%. Most unsupervised adaptation paradigms failed to achieve statistically significant reductions in error due to the uncertainty of the correct class. One method that selected high-confidence samples showed the most practical potential, although other methods warrant future investigation outside of a laboratory setting. The ability to provide supervised adaptation should be incorporated into any clinically viable pattern recognition controller, and unsupervised adaptation should receive renewed interest in order to provide invisible adaptation.