Construction of motion phase classification algorithm using ankle joint physical signals

Quantification of human muscular activity is crucial, especially the activity of the lower limb muscles, which plays a significant role in motor control and rehabilitation. Devices such as electromyographs, motion capture systems, and smartwatches have been used to achieve this purpose. These methodologies have been reported to contribute to the understanding of motor control mechanisms, elucidation of movement disorders, and optimization of rehabilitation interventions and programs among others. However, they present several challenges. This study proposed the selector circuit type LstpR-HMM, a universal algorithm for estimating movement states using the physical signals obtained from a device we developed to measure vibrations and angles around the ankle joint. The algorithm optimizes specificity and sensitivity by transforming the logistic function, which can thus prevent from the convergence to erroneous states due to transient noise. Moreover, it possesses interpretable internal mechanisms. The efficacy of this method was demonstrated by evaluating it on subjects with various physical characteristics who performed heel-raising exercises. The results thus highlighted the effectiveness of our proposed method.