Independently Coupled HMM Switching Classifier for a Bimodel Brain-Machine Interface

Abstract

Our initial attempt to develop a switching classifier used vector quantization to compress the multi-dimensional neural data recorded from multiple cortical areas of an owl monkey, into a discrete symbol for use in a single Hidden Markov Model (HMM) or HMM chain. After classification, different neural data is delegated to local linear predictors when the monkey's arm is moving and when it is at rest. This multiple-model approach helped to validate the hypothesis that by switching the neuronal firing data, the performance of the final linear prediction is improved. In this paper, we take the idea of using multiple models a step further and apply the concept to our actual switching classifier. This new structure uses an ensemble of single neural-channel HMM chains to form an Independently Coupled Hidden Markov Model (ICHMM). Consequently, this classifier takes advantage of the neural firing properties and allows for the removal of Vector Quantization while jointly improving the classification performance and the subsequent linear prediction of the trajectory.