Gender estimation according to Gait by using ellipse fitting and static body parameter approaches

In this study, by using Support Vector Machine (SVM) and Learning Vector Quantization (LVQ) classifiers, the issue of “gender estimation according to gait” is covered. The images used in the study are provided from the CASIA Gait Database. After the images are categorized according to gender, training and test data sets are constructed. In the next step, the gait images belonging to each person in the data sets are selected so that they complete a cycle (two footsteps), the remaining of the images are removed, and for each remaining array of images, feature extraction is carried out by using the ellipse fitting and static body parameter approaches together firstly in the literature. By giving the features extracted by using both of the approaches on the training dataset to SVM and LVQ classifiers, training processes are implemented and then, the features extracted from the test data by using the same approaches are given to these classifiers. After the classification processes, the average correct classification rates for SVM and LVQ are 100% and 90% respectively.