Research On Signal Processing And Pattern Recognition Based On The Gas Sensor Array

With the social developing rapidly, gas detection technology is becoming more and more important. For example, gas detection in flammable and explosive for mine, can effectively prevent gas explosion accidents. Gas detection around the industrial toxic gas tanks, can discover and report toxic gas leakage timely, so that can be the greatest degree to reduce the harm caused by leakage of toxic gases. The gas detection in volatility of vegetable and food, not only can judge how fresh vegetable and food are, but also can distinction which level the product is. For patients with the gas exhaled testing, can quickly realize diagnosis, which saving time and a lot of manpower, and it is very convenient, ... . All these examples show the practicality and importance of electronic nose detection technology, thus further development and promotion of electronic nose technology becomes an trend of gas detection technology.It is well known that, the hardware of electronic nose can not make a breakthrough in the short time. But software can achieve updates or improvement in short term, thus, most researchers began to focus on software algorithm of the electronic nose technology. The software algorithm can provide high recognition accuracy and low recognition time, which can compensate the error caused by hardware and environmental uncertainty. Pattern recognition algorithm is the key technology of electronic nose software. Data for algorithm must be prepared, the process is called feature extraction and selection. This paper also focus on the study of signal processing and pattern recognition algorithm about electronic nose. We disign a new style PID controller to improve the dynamic performance of data acquisition system, and put forward a new method to produce feature vectors, then put the feature vector into the designed pattern recognition system and realizes the classification finally.Main research work we have done as follow:Firstly, we analysed the current research progress of this field, and given the most comprehensive review about algorithm and application, it is very useful for some researchers who begin focus on this field.Secondly, we introduced several common signal processing and pattern recognition algorithm in the research of this field, and analysed the advantages and disadvantages of them.Thirdly, we researched how to use the truncated technology to get the data of sensor response. Our research based on two filter approximation algorithms, the optimal Hankel approximation with direct truncation algorithm approximate algorithm were studied. Simulation results show the feasibility of data truncation technology applied to data collection.Finally, we designed a new signal processing and pattern recognition algorithm, and the classification result show the superiority of the new method.