Classification of Human Gender from Sweat Odor using Electronic Nose with Machine Learning Methods

Abstract

Both human biological genders have the same hormone but at different levels. The difference in hormone levels makes the two genders distinguishable from several aspects. One of the things that are influenced by hormones is sweat. The odor of sweat is related to the apocrine glands found in human armpits. This experiment studied the classification of both genders based on daytime sweat in adult human armpits. The sampling method used an electronic nose (E-nose) system to collect the armpit sweat odor. The E-nose system sensor array consisted of seven sensors: TGS 822, TGS 2612, TGS 2620, TGS 826, TGS 2603, TGS 2600, and TGS 813. These sensors generate resistance ratio (Rs/Ro) values which are learned by the machine learning methods for classification and disease potential based on the volatile organic compound (VOC) in sweat. The study shows the male samples have higher amine gas than female samples, one of which is Trimethylamine (TMA). TMA is a compound that will be broken down into trimethylamine-N-oxide (TMAO), a factor to various cardiovascular diseases. The result achieved 94.12% accuracy in classifying human biological gender using principal component analysis (PCA) as the pre-processing method and support vector machine (SVM) as the machine learning method.