<p>Gas sensor is an important technique for food monitoring. However, its performance is often influenced by environmental temperature and humidity. In this study, a new gas sensor signal compensation method was proposed. The method was based on absolute humidity (AH), an important indicator related to the sensor response mechanism. But, the AH based model is computationally complex, which hinders its practical application. Hence, a simplified AH based model was established by eliminating the high-order terms with weak influence, generating an equation with no more than 3 orders and thus reducing computational complexity. First, the sensor signal drift characteristic was analyzed, indicating a strong correlation between the sensor signal and temperature-humidity. The results proved the feasibility of signal compensation. Then, the optimal model for each sensor was selected and its parameters were optimized based on the mean value and standard deviation. Finally, the performances of models before and after compensation were compared using MLP and SVM and five-fold cross-validation was employed. Moreover, the model was constructed based on the background environment, and directly applied in wine and banana samples. Both the MLP and SVM results showed a minimum 6% increase in classification accuracy and more than 82% of the classification accuracy was achieved. This verified good compensation effect and a certain level of generalization ability of the proposed method. These findings confirm that the proposed method is effective in compensating gas sensor signals for food detection under varying temperature and humidity conditions, demonstrating potential in practical applications.</p>