Skip to Main Content
Estimation of on-off timing of human skeletal muscles during movement is an ongoing issue in surface electromyography (sEMG) signal processing for relevant clinical applications. Widely used single threshold methods still rely on the experience of the operator to manually establish a threshold level. In this paper, a novel approach to address this issue is presented. Based on the generalized likelihood ratio test, the maximum likelihood (ML) method is improved with an adaptive threshold technique based on the signal-to-noise ratio (SNR) estimate in the initial time before accurate sEMG analyses. The dependence of optimal threshold on SNR is determined by minimizing the onset/offset estimate error on a large set of simulated signals with well-known signal parameters. Accuracy and precision of the algorithm were assessed by using a set of simulated signals and real sEMG signals recorded from two healthy subjects during elbow flexion-extension movements with and without workload. Comparison with traditional algorithms shows that with a moderate increase in the computational effort the ML algorithm performs well even for low levels of EMG activity, while the proposed adaptive method is most robust with respect to variations in SNRs. Also, we discuss the results of analyzing the sEMG recordings from the selected proximal muscles of the upper limb in two hemiparetic subjects. The detection algorithm is automatic and user-independent, managing the detection of both onset and offset activation, and is applicable in presence of noise allowing use by skilled and unskilled operators alike.