Authors : Noor Saffazura Ahmad Safuan, Marina Ismail and Nursuriati Jamil
Abstract: This analysis is to compare the PCA, KPCA, LDA, ICA, LBP and LDP technique and suggestion on suitable techniques for gait analysis on children and adult; for both normal and with impairment. This study reviewed journal and studys in the area of computer sciences. All studys were restricted to english publications only. Study that discussed on gait video analysis, gait recognition and feature extraction techniques has been analysed. Any study related to PCA, KPCA, LDA, ICA, LBP and LDP techniques were also included in this analysis. Document that >2009 based on date of publication were excluded from this analysis. The comparison of the six techniques has been made based on the researchers comments and suggestions in the study reviewed. The comparison has been divided into 3 categories which are the advantages, disadvantages and similarity. The 5 features has been chosen as techniques that are related to the gait parameters. We have found that all techniques can handle the spatio temporal data. Spatio-temporal data is a calculation based on gait parameter such as velocity, cadence, time, stride duration, stride length, step width. Based on the spatio temporal data, the techniques can map the data either in linear structure or nonlinear structure. Linear structure means the technique has map the data in a straight line while for nonlinear structure the techniques can map the data into non straight line or curvy lines. There are two types of feature which are local feature and global feature. Local features extracted salient point in the input domain while global features extracted from the complete gait image. This study can be enhance by comparing more techniques available and perform an experiment on those techniques to obtain more precise result.
Noor Saffazura Ahmad Safuan, Marina Ismail and Nursuriati Jamil, 2017. Feature Extraction Technique for Human Gait Video Analysis. Journal of Engineering and Applied Sciences, 12: 534-541.