Temperature correction to chemoresistive sensors in an e-NOSE-ANN system

The influence of the temperature coefficient of resistance in the chemoresistive response of inherently conductive polymer (ICP) sensors in the performance of an artificial neural network (ANN) e-natural olfactory sensor emulator (e-NOSE) system is evaluated. Temperature was found to strongly influence the response of the chemoresistors, even over modest ranges (ca. 2/spl deg/C). An e-NOSE array of eight ICP sensor elements, a relative humidity (RH/spl plusmn/0.1%) sensor, and a resistance temperature device (RTD/spl plusmn/0.1/spl deg/C) was tested at five different RH levels while the temperature was allowed to vary with the ambient. A temperature correction algorithm based on the temperature coefficient of resistance /spl beta/ for each material was independently and empirically determined then applied to the raw sensor data prior to input to the ANN. Conversely, uncorrected data was also passed to the ANN. The performance of the ANN was evaluated by determining the error found between the actual humidity versus the calculated humidity. The error obtained using raw input sensor data was found to be 10.5% and using temperature corrected data, 9.3%. This negligible difference demonstrates that the ANN was capable of adequately addressing the temperature dependence of the chemoresistive sensors once temperature was inclusively passed to the ANN.