Pseudo-online Muscle Onset Detection Algorithm with Threshold Auto-Adjustment for Lower Limb Exoskeleton Control

Abstract

Spinal cord injury (SCI) is one of the main disabling injuries affecting people worldwide. The use of a wearable robot in rehabilitation reduces physical exhaustion during training. Moreover, developing novel control paradigms that allow the wearable system to move in a more intuitively way, resembling patients’ intention, may enhance rehabilitation outcomes. In this paper, we present an algorithm that automatically detects users’ onset and offset of lower limb muscle’s contraction using electromyography (EMG) signals. The users’ intention was detected through a single threshold algorithm. The algorithm auto-adjusts the threshold according to individual EMG characteristics. The algorithm was tested with EMG data from seven healthy subjects performing ankle and knee flexion/extension movements and correctly detected the intention of movement in approximately 96% of cases in an average time of 147.15 ms. To conclude, this algorithm can potentially be explored in future approaches to enable real-time triggering and control of a wearable exoskeleton to be used in the rehabilitation setting.