DTW 전처리를 통한 SVM 분류기 설계

This paper presents a SVM classifier design by DTW preprocessing. To evaluate the performance of proposed classifier, data from underwater noise database was used. Database is comprised of biological transient noises, mechanical transient noises and so on. We design a classifier to classify particular mechanical noises. MFCC feature vector that represents the characteristics of the transient data is used to the SVM-BTA classifier based on the SVM. DTW result is used for design the SVM classifier. As the simulation results, the performance was improved SVM using the DTW result than using categorization of database.