Bio-Inspired Background Suppression Technique and its Implementation into Digital Circuit

Background

The adaptation is a common biological function in a sensory system, which allows the sensory system to reduce the sensitivity to previously detected stimuli, and then to improve the selectivity to a novel and interesting stimuli. In this chapter, we focus on the adaptation in the olfaction for developing bio-inspired signal processing methods for gas/odor sensing.

Neural computation models for olfactory adaptation or mixture segmentation have been proposed in various researchers (Figure 1). For example, Wang, Buhmann, and Malsburg (1990) proposed a neural model of pattern segmentation based on a neural network of associative memory, as shown in Figure 1 (a). They employed a neural network of oscillatory units linked with Hebbian connections to perform temporal segmentation of the stored patterns. Alternating bursts of activity induced by self-inhibition creates a spatio-temporal pattern that sequentially extracts the components of mixture.

Figure 1.

Neural computation models for olfactory adaptation or mixture segmentation. (a) Mutually connected oscillatory units (Wang, Buhmann, & Malsburg, 1990), (b) hierarchical layers of neural network (Hendin, Horn, & Hopfield, 1994), (c) hierarchical neural network model of olfactory cortex (Oyamada, Kashimori, Hoshino, & Kambara, 2000), (d) bulb-cortex model with feedforward and feedback paths (Li & Hertz, 2000), (e) KIII model (Yao & Freeman, 1990)