Abstract: Moldy foodstuff with mildew toxin are harmful to people and animals. To provide a simple and objective solution to identify whether foodstuff are moldy, an artificial olfactory system which consists of gas sensor array, gas tube system, signal adjusting system, data collecting system and pattern recognition system was designed. In order to improve veracity of identification, three eigenvalues were used in collecting data, one is the maximal response point, the other two are the left and the right point at some distances of the maximal response point and the sample eigenvalues were trained with the triplex-optimized BP Neural Network. After testing, the trained samples and the tested samples can both be identified correctly, which indicates that the artificial olfactory system is veracious and effective.