An embedded probabilistic neural network with on-chip learning capability

Abstract

An embedded system capable of recognizing biomedical signals reliably is important for fusing sensory data of portable or implantable microsystems in biomedical applications. This paper presents the digital VLSI implementation of the probabilistic neural network, called the Continuous Restricted Boltzmann Machine (CRBM), which is able to cluster or to classify sensory data of an electronic nose. The learning algorithm of the CRBM is also realized on the same chip, such that the CRBM system is able to optimize its parameters automatically, or to compensate for sensory drifts by on-line learning.