在建立人工嗅敏实验系统的基础上针对使用BP网络处理气敏阵列信号时出现的种种不足,引入较之优越的ML算法,并提出用非线性函数及SOM网络对气敏样本进行预处理,用分段训练的方法加速学习过程,最后得到在速度和精度上更接近实用要求的新的嗅敏信息处理合成算法. Based on the establishment of the experimental system of artificial sniffing and aiming at the disadvantages of using the BP neural network to process the gas sensor array signals, the superior ML algorithm is introduced and the preprocessing of the gas sensitive sample with the non-linear function and the SOM network is proposed. The method of segmented training is used to speed up the learning process, and finally a new synthetic algorithm of sniff sensitivity information processing is obtained which is closer to the practical requirement in speed and accuracy.