Robust supervised classification of motor unit action potentials.

摘要: A certainty-based classification algorithm is described, which comprises part of a clinically used EMG signal decomposition system. This classifies candidate motor unit action potential (MUAP) to the train (MUAPT) that produces greatest estimated certainty, provided this maximal certainty above given threshold. The iterative, such with assignments are made increases each pass through data, and it has specific stopping criteria. performance sensitivity (to assignment threshold) Certainty an iterative minimum Euclidean distance (MED) compared by classifying sets MUAPs detected in real concentric needle-detected signals, using range thresholds for algorithm. With regard MUAP error rates, consistently provides better mean results and, more importantly, less variable than MED can provide rates 80.8 1.5%, respectively, maximum rate 3.2%; 80.3 3.3%, 6.5%. relatively insensitive threshold used, differentiate between similarly shaped from different MUAPTs, make correct classifications despite biological shape variability, background noise non-stationarity.