Recognition Feature Extraction of Pernicious Gases in Piggery Based on Wavelet Transform and Genetic Algorithm

Electronic-nose was used on detecting the pernicious gases in piggery. The feasibility of improving the electronic-nose recognition model by using a feature extraction method with wavelet transform and genetic algorithm (GA) was discussed by aiming on cross-sensitivity of gas sensors. The experiment result shows that, the feature samples extracted by the new method, these samples were inputted in back-propagation neural network (BPNN) could greatly enhance the learning speed of the BPNN to a certain recognition right-rate compared with principal component analysis (PCA). The quantitative recognition error of the mixture of ammonia and hydrogen sulfide gas samples was also reduced on the net so that the identification precision was enhanced.