Robotic technologies are progressively gaining considerable importance in motor rehabilitation. In this context, the development of non-invasive man-machine interfaces has a significant role. Among other physiological signals, surface EMG is of paramount importance. However, the detection of surface EMG signals in rehabilitation is currently based almost exclusively on a single or a few electrode pairs. In the last decade, some important limitations of this approach emerged. This work reviews the more recent knowledge about issues in the use of simple bipolar system when the objective is to estimate muscle activation and highlight the benefits provided by multi-channel surface EMG to muscle activity estimation.