Quantification of Individual Gases/Odors Using Dynamic Responses of Gas Sensor Array With ASM Feature Technique

This paper is a continuation of our previous work in which a new feature technique called average slope multiplication (ASM) was proposed to classify the individual gases/odors using dynamic responses of sensor array. The ASM method is used to quantify the individual gases/odors in this paper. Back propagation algorithm based two different neural network architectures (NNAs) called NNA1 and NNA2 are used to assess the ability of the ASM technique for quantification. The proposed method thus utilizes the newly developed feature method in the first stage and the specially designed neural quantifiers in the next subsequent stages. The ability of the proposed method has been insured by applying it on the published dynamic responses of the thick film gas sensor array. When the raw data were directly fed to the neural quantifiers, the results were 69% and 63% accurate for NNA1 and NNA2, respectively. The principal component analysis preprocessed version of raw data provided 74% and 67% quantification accuracy with the aforementioned architectures respectively. The performances of the ASM data were found to be 100% using both the network architecture without need of further preprocessing, with relatively less number of epochs and without any hidden layer. Thus, the proposed method can be utilized in electronic nose for classification/quantification purpose.