Unsupervised detection and classification of motor unit action potentials in intramuscular electromyography signals

A computationally efficient and unsupervised algorithm for the detection and clustering of motor unit action potentials (MUAPs) recorded in a single-channel Intramuscular Electromyography (EMG) is presented. The detection of MUAPs is performed with a modified version of the multiresolution Teager energy operator (MTEO). The unsupervised clustering of action potentials is achieved by applying a combination of label and template matching techniques. The proposed algorithm reduces the partial superimposition of MUAPs with a new MTEO based analysis method. The computational speed of the method is also improved by using the principal component analysis (PCA) in order to reduce the number of templates and fiducial point detection and consequently to decrease the correlation computation load. The performance of the algorithm is examined on several intramuscular EMG recordings of different healthy and diseased muscles such as the posterior cricoarytenoid and tibiliasis anterior.