Classifying Odor Concentration With An Electronic Nose

In this paper, research actuality of electronic nose and its application field are introduced. The advantages of electronic nose in odor detection are analyzed. The hardware system of electronic nose is introduced in detail. The odor signal processing methods based electronic nose is researched in full-scale and systematically. At last, some tested data are used to testify the effect of all the processing methods.Some commercial odor sensors and their performance are introduced and compared. Then the three key parts of electronic nose: sensors array, signal amplification circuit and AD board are also introduced.The output of electronic nose is from several sensors. Each sensor's output is the superposition of the sensor's response to all tested gases. So electronic nose signal is a kind of complicated high dimensional signal. In this paper, some signal processing methods which suit for electronic nose are studied and discussed. The principal component analysis (PCA) is used in reducing the dimension of E-nose signal. The structure and arithmetic of artificial neural network such as the BP network, the SOM network and fuzzy network are introduced. The effect of independent component analysis (ICA) in picking up the character of E-nose signal is also studied.In this paper, we presented a gas classification model which based on synthetic consideration of all the processing methods. Some experiments are performed to verify the effect of PCA and ICA. The results of the three neural network models are discussed and the best one is found out.At last, the prospect of electronic nose is expected and some ideas for later research are given.