DTW를 이용한 보행 분석 시 warping path 특성

Abstract

The purpose of this study was to investigate the characteristics of the warping path on gait patterns using the DTW algorithm. Participants in this study were consist of 20’s age of 10males and 10 females. Gait data obtained by a 3D motion analysis system analyzed using the DTW program made by Matlab’s functions. The analysis was performed 120 times including normal walking and abnormal walking. The length of the warping path relative to the reference sequences, and deviations was analyzed for the warping path calculated by the DTW program. The statistical method used in this study was a correlation, independent t-test, one-way ANOVA, and regression. The statistical significance level was  =.05. In this study, the correlation coefficient of Pearson was =.816 between the warping path length and the deviation, and this relationship was verified by regression analysis. The statistically significant difference between normal walking and abnormal walking appeared in the warping path length and in the deviation. The analysis according to walking phases showed a statistically significant defference between the support phase and the swing phase. In conclusion, if the length of the warping path is greater than about 127% compared to the diagonal length between two sequences, normal walking can be suspected, and if the mean of the warping path length deviations is greater than about 5.8, normal walking can be suspected also.