Summary
International Symposium on Nonlinear Theory and its Applications
2017
Session Number:A3L-E-2
Session:
Number:A3L-E-2-3
Time Series Classification with New Similarity Measure: an Application for Automatic Detection of Driver's Distraction
Basabi Chakraborty, Sho Yoshida,
pp.326-329
Publication Date:2017/12/4
Online ISSN:2188-5079
DOI:10.34385/proc.29.A3L-E-2-3
Summary:
Classification or grouping of time series data is now increasingly needed to solve various real life problems. As time series data is huge, a proper representation method and an efficient similarity measure are important factors for the success of any clustering or classification method involving time series data. Though a lot of research has already been done in this line, dynamic time warping (DTW) seems to be the most common method used for measuring similarity of two time series data. Though classification accuracy of time series classification with DTW is quite satisfactory, computational cost is also very high. In this work, newly proposed measures by the authors have been used for time series classification problem. Publicly available benchmark data sets as well as time series data from a real life problem of detecting driver's distraction with cognitive load are used for classification with the proposed measures. The comparative effectiveness of the proposed measures over DTW has been examined by the experimental results.