GC–MS characterization of body odour for identification using artificial neural network classifiers fusion

摘要: Abstract The focus of the present study is human body odour recognition by analysis information about chemical compounds identified in their gas chromatography–mass spectrometry (GC–MS) chromatogram. Artificial neural network (ANN) technique implemented current study, has been comprehensively used for classification and regression tasks numerous applications. experimental data set includes intensity characteristics (peak height, peak area, ratio area height) several detected GC–MS chromatogram twenty samples (from four persons), two non-body samples. raw transformed with logarithmic scaling, principal component (PCA), kernel (KPCA) search better features extracting. After preprocessing data, feed forward back-propagation (BPNN) discrimination samples, as well to an individual. Although ANN classifier optimized number neurons, training algorithms, result unstable unsatisfactory (maximum correct rate 78% minimum 44%). To improve stability accuracy results, fusion approach attempted. Eight different weighted unweighted decision schemes have recognition. Amongst them simple vote (SWV), quadratic best worst (QBWWV), (BWWV) outperform 100% class outcomes, compared a single classifier.