Gait Signals Classification And Comparison

Use of wireless signal technology in sensing of human gait activity is a satisfactory example of device-free sensing and effective in medical science to detect human motion-related diseases. Some prior research showed some potential detecting process of human walking gait from wireless channel information (WCI) using wireless signals. In this paper, we present comparison of three popular features reduction methods such as principal component analysis (PCA), kernel principal component analysis (KPCA), and linear discriminant analysis (LDA) using three classifications methods, support vector machine (SVM), k-nearest neighbor (k-NN), and decision tree (DT) in an absolutely equivalent situation for identifying walking gait signals. The analysis was carried out on the WCI-based dataset where dataset was divided into four classes (normal gait, small gait, fast gait, and turn gait). Using dataset with the combination of methods (features reduction and classification), experimental results shows that all the combinations of PCA, KPCA, and LDA with three classifications achieve an average accuracy of gait identification is accordingly 86%, 79%, and 95%.