Rapid odor recognition based on reliefF algorithm using electronic nose and its application in fruit identification and classification

Recognition time-consuming and accuracy are two important parameters for an electronic nose (E-nose) system. Most reported E-nose systems were based on the steady-state response or the entire response process of gas sensors, which result in a relatively long time for the recognition process. In addition, Principal Component Analysis (PCA), the most widely used method in the field of odor recognition, often fails to extract the key features for achieving recognition tasks. This usually reduces the recognition accuracy of the E-nose system. In order to overcome the above problems, this paper proposed a novel odor recognition method for E-nose system based on the start stage of sensor response and ReliefF algorithm, and applied it to identify and classify three categories of fresh and spoiled fruits (apple, pitaya, and tribute citru). The results showed that extracting features only from the start stage of sensor response can greatly shorten the odor recognition time. Compared with the traditional PCA method, ReliefF can select the key features more efficiently and thus improve the recognition accuracy of the E-nose system.