Single feature to achieve gas recognition: Humidity interference suppression strategy based on temperature modulation and principal component linear discriminant analysis

Abstract

Metal oxide semiconductor (MOS) sensors have been broadly employed for gas detection. However, the distinctive chemical detection principle of MOS sensors renders them susceptible to interference by humidity. This paper proposes a novel method for suppressing humidity interference. This method possesses advantages, including rapid recognition, low power consumption, and the capability to achieve recognition using a single feature. Temperature modulation technology was used to record the sensor's resistance variation, and the response features of the obtained dynamic response signal are affected by both humidity and the measured gas. The suppression of humidity interference is achieved through the investigation of the response features of humidity and the measured gas. The response features of humidity and gas are amplified by column normalization. Principal Component Linear Discriminant Analysis (PC-LDA) is utilized to acquire humidity and gas information within data. Using ethanol as the primary experimental subject, suppression of humidity interference is achieved through a single feature (PC-LD2). Quantitative recognition of relative humidity (RH) can be achieved using PC-LD1. Multiple types of waveform and sensor were used to demonstrate the generalization of this method.