PROBABILISTIC INFERENCE-BASED CLASSIFICATION APPLIED TO MYOELECTRIC SIGNAL DECOMPOSITION

A new probabilistic inference based technique (IBC) for the classifica tion of motor unit action potentials (MUAP's) is presented. This new t echnique discovers statistically significant relationships in the data and uses these relationships to generate classification rules. The te chnique was applied to the classification of MUAP's extracted from sim ulated myoelectric signals. Its performance was compared to that of cl assical template matching algorithms (TBC) applied to the same data. U sing 32 time samples as features to represent the MUAP's it was found that the IBC based technique performed significantly better ( p < 0.00 5) than the TBC algorithms (83.0 +/- 2.6% versus 78.1 +/- 2.8% peak co rrect classification performance). As the size of the training set was reduced or as increasing numbers of random classification errors were introduced into the training data, the performance of the IBC and TBC techniques declined similarly. IBC performance remained superior unti l very small training sets (< 30 MUAP's per motor unit) or training se ts with large numbers of errors (> 50%) were used. Because the probabi listic inference technique can utilize nominal data it has the potenti al to use declarative problem domain knowledge which conceivably could improve its performance.