A Blind Automatic Detection of Surface EMG Bursts using an Adaptive Mask

The analysis of the electromyography signal (EMG) can be exploited in several scenarios, from clinical practice to rehabilitation or sport training. However, the identification of each EMG burst often relies to a visual inspection of an expert. In this context, this work aims to present a new method for the automatic detection of EMG bursts. The proposed method bases its functionality on the creation of a mask for assessing peaks amplitude and duration. Two versions of the algorithm are presented: offline modality and real-time simulation mode. The performances of the algorithm were tested on the isometric contraction of brachial biceps. Applied to thirteen subjects, the offline algorithm shows encouraging results with an F1score of 0.94. This result is in line with similar findings presented in the literature, but applying threshold-based methods that depend on the muscle of interest or on the signal processing. The application of this method shows a statistical significant increasing of the signal-to-noise ratio (SNR). The assessment of the performance of the real-time simulation modality on eight subjects, highlights the influence of the training time in a correct identification of the successive bursts.