Bayesian Inference using Spike Latency Codes for Quantification of Health Endangering Formaldehyde

摘要

Recently, the exposure to formaldehyde has appeared as a major concern since it is listed as a human carcinogen. Conventional methods for its long-term monitoring are not feasible due to their high operational cost, long analysis time and the requirement of specialized equipment and staff. In this paper, we use an electronic nose, containing an array of commercially avail-able Figaro gas sensors, to estimate formaldehyde concentration. A hardware friendly bio-inspired spike latency coding scheme has recently been employed for gas classification by using relative time between spikes. We use this scheme to estimate formaldehyde concentration by utilizing absolute spike timings. However, there is no straightforward relationship between the spike latency and the formaldehyde concentration. Instead, stochastic variability in the sensor array response, corresponding to repeated exposure to the same formaldehyde concentration, implies that latency patterns of the sensor array encode probability distribution over the formaldehyde strength. We use a Bayesian inference approach to estimate the formaldehyde concentration, and its performance is successfully validated by acquiring data for formaldehyde with our sensor array at twenty different concentrations in the laboratory environment.