Research On Processing Methods Of Gas Spectrogram In Electronic Nose Based On Light Absorption Sensing

As a typical product of the development of artificial olfactory technology, electronic nose has been greatly developed in the past twenty years, and there are already some commercial products. But, the existing commercial electronic nose is still not up to the level of people's expectation. And the basic reason is that the performance of the gas sensing system in existing electronic nose is not perfect. Specifically, sensors in electronic nose cannot be as stable as imagined. They are easily influenced by interferences of environmental temperature, humidity and atmospheric pressure variation. Besides, the number of sensing units in existing electronic noses is small, which results in a limited range of sensing. In view of this, a gas sensing system based on broadband light absorption was proposed to improve the performance of the gas sensing system in electronic nose. Its main advantages are as follows: sensing unit in the gas sensing array based on broadband light absorption not only has cross sensitivity but also has wide response, and their number can reach to tens of thousands more, which is far more than that of any existing electronic nose. Besides, gas sensing based on light absorption also has other advantages like high speed, high sensitivity and long life, etc..Although gas sensing based on broadband light absorption has great potential for carrying the core sensing task of electronic nose, there are many problems need to be solved. Thereinto, the key problems includs: how to build a gas sensing system with large-scale sensing array based on broadband light absorption. And how to process data to get the fault-tolerant, stable and representative response according to the built system. Since in the electronic nose, qualitative and quantitative analysis of gas both depend on the pattern recognition, which is different from the traditional gas analysis instruments, the existing data processing methods are not targeted, so that the gas information cannot be obatined adequately. For these problems, the main research contents and contributions of this thesis are as follows:(1) The design, construction and tests of the proposed gas sensing system based on broadband light absorption for electronic nose.To improve the sensing performance of electronic nose, a gas sensing technique based on broadband light absorption was proposed. To verify the feasibility of using it to carry the core sensing task of electronic nose, mechanism of the gas sensing technique based on broadband light absorption was studied according to absorption spectroscopy firstly. Then theoretical model of the gas sensing response in electronic nose was formed. And the gas sensing system based on broadband light absorption was designed and constructed. Based on the constructed gas sensing system, responses both consistent with the theoretical model and data from HITRAN database were obtained.(2) The baseline drift suppression technique of spectrometer based on independent component analysis(ICA).In the constructed system, since CCD-based spectrometer was used, the baseline drift interference was introduced into the sensing response by the spectrometer, which mainly appeared as a phenomenon of light absorption of gas being weakened or even covered up by the interference. This reduces the accuracy and sensitivity of the system and makes the sensing performance worse. For this issue, firstly, the baseline drift data were analyzed and the characteristics and law of the interference brought by the baseline drift of the spectrometer were obtained. Then, according to the characteristics and law of the drift, a baseline drift suppression technique for spectrometer was proposed. Specifically, the proposed technique mainly includes four steps: data preprocessing, dimension reduction using principal component analysis(PCA), sources separation using independent component analysis(ICA) and reconstruction after interference suppression. At last, experimental results showed that, the proposed method suppressed the interference effectively, and more superior compared with the classicalexisting method.(3) A novel feature extraction method for gas spectrogram based on Global Variance Statistics(GVS).In the constructed system, due to the large scale of gas sensing array, the data process load is correspondingly enormous. Specifically, the number of responses obtained at the same time through the sensing array is large, which is several to dozens of times over that of the existing electronic nose, so the useful information extraction is a problem. In the view of this and according to the characteristic of electronic nose that the synthetic feature acquisition and recognition is the core, a novel feature extraction method for gas spectrogram based on Global Variance Statistics was proposed, so that the useful information can be extracted the from massive data with higher correct rate. Besides, the theoretical proof of the validation of the proposed feature extraction and fault tolerance mechanism was also given. At last, experimental results show that the proposed method was effective and superior to the existing feature extraction methods.(4) The broadband spatial heterodyne spectroscopy interferogram processing method based on image feature extraction.To introduce the gas sensing technique based on broadband light absorption into electronic nose, gas sensing array with massive parallel sensing units needs to be constructed. However, it is difficult to meet the requirements of both high resolution and a wide spectrum only in the grating spectrum method. In view of this, broadband spatial heterodyne spectroscopy was brought into the proposed gas sensing technique to realize the spectra generation and detection, which provides feasible basis to construct a gas sensing array with large-scale and wide response range. However, in traditional methods, the inverse Fourier transform of the broadband spatial heterodyne spectroscopy interferogram is needed to reconstruct the spectrum of the input signal light, and then the gas features are extracted from the reconstructed spectrum. This process is not only complicated but also has a large amount of computation. In view of this, analyze the characteristic of electronic nose that synthetic feature acquisition and recognition are the core. And if there is a one-to-one correspondence between the interferogram and the input signal spectrum, i.e., the interferogram is unique. The interferogram feature can be used as the gas feature during the pattern recognition in electronic nose. Hereinto, the mathematical model of the interferogram of broadband spatial heterodyne spectroscopy was studied, and the method to verify whether it is it a one-to-one correspondence between the input signal spectrum and the interferogram obtained was proposed. Simulation results showed the validness of the proposed method. Besides, performance of different kinds of texture feature extraction methods on extracting features from broadband spatial heterodyne spectroscopy interferogram was compared, and the types of suitable feature extraction methods in different situation were obtained.