An Empirical Analysis on Coincident point of 1D and 2D data for Behavioural Pattern Detection

Abstract

The present research paper illustrates thorough analysis on the coincident point values of 1D and 2D data. Here 1D and 2D data are the human-speech signal and human-gait image respectively considered for the detection of behavioural pattern. The work has been carried out by extracting four features from each data. The mapping process through coincident point theorem and soft-computing has been done. The video data is separated using adaptive-lifting-scheme of wavelet transform, which works on three factors: separation, prediction and updation. Pre-processing operations on 1D and 2D data have been done with proper filtration, enhancement, loss-less compression, segmentation and extraction of features using statistical and soft-computing methods. The video data is captured when the subject is talking while walking. Five varieties of natural languages: Oriya, Hindi, Bengali, Chhattisgarhi and English, have been used for the mapping with the coincident point of 1D and 2D data. The four 1D data extracted features are: speech duration, speaking rate, pitch and speech momentum. Similarly, the four 2D data extracted features are: step length, energy or effort, walking-speed and gait momentum. The coincident point on these two 1D and 2D are computed using best-fit measurement and hence the classification process has been done for the behavioural pattern detection. In the present work, 50 male subjects with varying age’s are used for an empirical analysis on the coincident point of 1D and 2D data and found that the performance of behavioural pattern detection has improved.