ABSTRACT
This paper describes a novel method to identify the composition of a binary gas mixture using a temperature-modulated gas sensor and a wavelet transform. Previous work has been reported that combines a wavelet technique together with neural networks to perform single component analysis. In this paper we report on a simpler method employing just the appropriate wavelet coefficients to identify concentrations of CO, NO2 in a binary mixture. The measurement system comprises precision analogue circuitry both to generate temperature modulations of an integrated platinum resistive heater, and to monitor the changes in the resistance of a metal oxide coating. Results pertaining to CO and NO2 gas mixtures are reported with a typical average operating temperature of 330 C° and a thermal oscillation of ± 80 C° at a frequency of 50 mHz. The wavelet coefficients of these gases were selected with the highest separation. The wavelet functions were found using MATLAB running on a Unix system. However, wavelet coefficients can be extracted by using a small number of additions and multiplication programmed in C, which subsequently can be run on an 8-bit microcontroller. The reduction of software overhead, coupled with the speed of data extraction from the sensor, is of importance in the production of a low-cost, palmtop gas monitor.
