[19a-212B-1~10] 1.3 Novel technologies and interdisciplinary engineering

△ [19a-212B-10] Signal Processing Method for Gas Sensor System Based on Transfer Function Toward Pump-Free Artificial Olfactory system

〇Gaku Imamura^1,2, Genki Yoshikawa^1,2,3, Takashi Washio⁴ (1.NIMS MANA, 2.NIMS CFSN, 3.Univ. Tsukuba, 4.ISIR Osaka Univ.)

Keywords:gas sensor, signal processing, machine learning

By combining a concept of control theory with signals processing of gas sensors, we have developed an analytical method to extract signal features that are independent of gas flow control, leading to a pump-free artificial olfactory system. In this study, we measured odors of five samples with a membrane-type surface stress sensor (MSS) by manually wagging the sensor chip in the odors of the samples. Based on the obtained time-series data of the sensor responses, we developed a classification model for the samples by using machine learning algorithms, resulting in a high classification accuracy. This study leads to a practical artificial olfactory sensor system that can adapt mobile devices.