Recurrence Plots-Assisted Deep Learning Based Metal Oxide Semiconductor Sensor Array System

Herein, a deep learning feature engineering approach for time series transformation of sensing responses is proposed to improve the ability of gas sensor array system to recognize gases. By transforming gas responses into recurrence plots, we extracted additional features from the gas response data, thereby enhancing the system's recognition capability. We process the gas sensing dataset with deep learning algorithms, the recognition accuracy of recurrence plots methods is significantly improved compared to methods based on time series processing from 84.94% to 97.85%, achieving high accuracy recognition of five gases.