HMM classifier using biophysically based CMOS dendrites for wordspotting

摘要

We explore the co-relations between Neural systems, CMOS transistors and Hidden Markov Models(HMM). We have built a computational model, implementing an HMM classifier that was built using biophysically based CMOS dendrites for wordspotting. The system was implemented on a reconfigurable analog platform. The system thus realized, was found to have high computational efficiency. We discuss the implications of such a computational model. We will also discuss how analog systems can effectively model biological systems, considering benefits both in terms of cost and power dissipation.