Muscle Deformation Sensing for Swimming Mode Identification and Continuous Phase Estimation With Two-Stage Network

Accurate recognition of human motion modes and continuous phases is crucial to exoskeleton control to provide proper assistance. However, harsh underwater environments severely restrict the study on swimming motion monitoring, where existing researches either focus on a single swimming mode or discrete phases, limiting underwater exoskeleton control. To address this limitation, this paper develops a two-stage network consisting of one mode classifier (1<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">st</sup> stage) and four phase regressors (2<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">nd</sup> stage), where muscle deformation features are used instead of traditional joint kinematics. Swimming tests are conducted on nine subjects with four modes at three frequencies. Effectiveness of the proposed method is justified by mode identification accuracy of 99.72% and phase estimation error of 3.92%, where the error is 52.89% smaller than that in the traditional time-based estimation method. This paper is the first to simultaneously recognize the swimming mode and the continuous phase, which is valuable to adapt the smooth exoskeleton assistance to harsh underwater environment and multimodal motion scenarios.