Abstract—The study aims to develop an state estimation framework for the detection of stance and swing phases of gait cycle during walking on a perturbed platform, i.e. a slackline. We use Support Vector Machine (SVM) classifier for the detection of stance and swing phase in gait cycle. Surface Electromyography (EMG) data was recorded from nine different muscles in the lower extremity from five healthy subjects. The proposed structure utilizes the hypothesis of Muscle Synergies (MS) where the movement intent is modelled as hidden state of the state space framework. We employ time domain modeling of the neural drive that excites the task-dependent muscles. To cater for the naturally existing physiological bounds (the non-negative muscle activations), the state estimation process is executed using a constrained form of the Kalman filter. Principal Component Analysis (PCA) is used for dimensional reduction of reconstructed EMG signal. We evaluated performance of SVM classifier, and the detection accuracy was later enhanced with post processing. Our preliminary experimental results demonstrate a reconstruction and classification accuracy greater than 95%. 

Index Terms—walking over slackline, electromyography, neural drive, muscle synergies, Kalman filter, Principal Component Analysis, Support Vector Machine

Cite: Safi Ullah, Kamran Iqbal, and Rajat Emanuel Singh, "Muscle Synergies Based Gait Phase Classification during Kinematically Constrained Walking on Slackline," International Journal of Electronics and Electrical Engineering, Vol. 8, No. 3, pp. 47-52, September 2020. doi: 10.18178/ijeee.8.3.47-52