Foot Gesture Recognition Through Dual Channel Wearable EMG System

Abstract

Surface electromyography (sEMG)-based hand gesture recognition (HGR) is finding its application as a general-purpose wearable human machine interface (HMI). However, the same is rarely explored for foot gesture recognition (FGR). The FGR is interesting as a hands-free controller, e.g. in case of an electric guitar player who controls the musical effects by foot pedals. Yet, the FGR is challenging, since general leg movements such as walking can be classified as foot gestures. In this paper, the application of a minimalistic sEMG wearable footband is demonstrated as a hands-free HMI. With only two channels and a support vector machine (SVM) classifier, the system is able to classify five foot gestures. In addition, we added a locking/unlocking mechanism and also controlled by one of the gestures, which eliminates undesired gesture classification during general leg movements. We show that the gesture-controlled locking feature is robust, and system does not unlock during walking, jumping, climbing the stairs, and similar. This paper covers the system design, the real-time classification algorithm, the selection of the target muscles, and experimental results. Sessions independence, accuracy, and robustness of the device are tested in a total of 18 sessions with three volunteers having different usage skills. As case study, the system was interfaced with a Window 10 application developed for controlling musical effects with the foot wearable band as a hands-free alternative for the DJ mixer equipment.