Motion Pattern Recognition Using EMG Signals

If we can decode human motion intentions from EMG signals, it can be applied to control various devices. For this purpose, it is necessary to estimate the correspondence between measured EMG signals and the operator's motion intentions using machine learning. Various classification models have been used for EMG-based motion pattern recognition, but most of these are based on general-purpose machine learning methods and do not sufficiently consider the generation process and probabilistic characteristics of EMG signals. In this research, we are working on constructing motion pattern recognition algorithms by applying research on stochastic generative models of EMG signals. Through this approach, we aim to achieve higher-precision motion pattern recognition by realizing learning and inference that utilizes the properties of EMG signals themselves, rather than relying on general-purpose machine learning methods that have been commonly used for EMG signal classification. Main research features: - Motion pattern recognition based on stochastic generative models - Recognition of unknown compound motions through pseudo-data generation