Spike Latency Coding in Biologically Inspired Microelectronic Nose

Abstract

Recent theoretical and experimental findings suggest that biological olfactory systems utilize relative latencies or time-to-first spikes for fast odor recognition. These time-domain encoding methods exhibit reduced computational requirements and improved classification robustness. In this paper, we introduce a microcontroller-based electronic nose system using time-domain encoding schemes to achieve a power-efficient, compact, and robust gas identification system. A compact (4.5 cm × 5 cm × 2.2 cm) electronic nose, which is integrated with a tin-oxide gas-sensor array and capable of wireless communication with computers or mobile phones through Bluetooth, was implemented and characterized by using three different gases (ethanol, carbon monoxide, and hydrogen). During operation, the readout circuit digitizes the gas-sensor resistances into a concentration-independent spike timing pattern, which is unique for each individual gas. Both sensing and recognition operations have been successfully demonstrated in hardware. Two classification algorithms (rank order and spike distance) have been implemented. Both algorithms do not require any explicit knowledge of the gas concentration to achieve simplified training procedures, and exhibit comparable performances with conventional pattern-recognition algorithms while enabling hardware-friendly implementation.

Interaction of cellular and network mechanisms for efficient pheromone coding in moths

“…The MGC of M. sexta moths could, therefore, make use of latencies to implement a dual coding scheme and resolve the apparent contradiction between quantitative coding (sensitive to changes in concentration and invariant to changes in proportion) and qualitative coding (sensitive to changes in proportion and invariant to changes in concentration). As an extension to this work, the proposed latency coding scheme provided a direct input for designing bio-inspired data analysis methods for artificial olfaction in electronic noses (42). In latency coding, one considers the neurons as analog to delay converters: the most strongly activated neurons tend to fire first, whereas more weakly activated cells fire later or not at all.…”

Probabilistic Rank Score Coding: A Robust Rank-Order Based Classifier for Electronic Nose Applications

“…This coding is successfully reported in the somatosensory [13], visual [14], auditory, [15] and olfactory systems [16]. This coding has been mimicked in rankorder based classifiers [17]- [19] for gas identification with electronic nose system. A logarithmic spike time encoding technique is used in these classifiers to generate a latency based spike sequence from the response vector of the sensor array.…”

“…Motivated by these findings, rank-order based classifiers have been developed by translating the response vector of a sensor array into a latency based spike sequence [17]- [19]. These coding schemes include 1-D spike rank-order, 2-D spike rank-order and glomerular latency coding.…”

Towards biological plausibility of electronic noses: A spiking neural network based approach for tea odour classification

“…Various researchers have attempted to represent the gas sensor response into spikes through schemes such as rate [5] and latency coding [2,[9][10][11]. Spike encoding of the metal oxide semiconductor response attempted by Hsieh et al [12], Yamani et al [2] and Chen et al [11] use the percentage resistance change and the logarithm of peak resistance which do not take into account the temporal information of the sensor response which in fact is the essence of EN data. These steady state methods take into account only stationary information of the sensor response and all the information related to the adsorption/desorption kinetics of the sensors and molecular dispersion of the analytes is lost [13].…”

“…In contrast to the previously described methods of sensor response encoding which encodes only the steady state response [2,11,12], we have been able to capture a wide range of sensor response through RFs distributed temporally over the entire sniff cycle. Additionally, methods which tend to capture the temporal response in spikes use rate-codes thereby generating dense spike patterns [4,5].…”