Detailed analysis of motor unit activity

We have developed a method for decomposing EMG signals into their constituent motor unit potentials (MUPs) and their firing patterns. The aim of the system is detailed analysis of motor unit variability. In the first phase of the decomposition, the EMG signal is separated into segments containing MUPs activity. The segments are identified by calculating the variance in a time window. The segments are then clustered by a minimum spanning tree method. This analysis leads to a partition consisting of clusters containing isolated MUPs and clusters composed of superimposed MUPs. The number of segments in a cluster is used to detect potentials from only one motor unit. From each of these clusters a template is selected. The clusters containing superimposed MUPs are analysed by a recursive algorithm. The cross-correlation between superimposed MUPs and a template are computed and time shifts with high correlation are detected. The template is subtracted for each of these time shifts, and the residual segments are processed by a subsequent pass through the algorithm. The output of the decomposition algorithm provides information about recruitment and firing rate of individual MUs