PENGENALAN SUARA PEMBICARA MENGUNAKAN METODE DYNAMIC TIME WARPING

Abstract

The world of technology is vast and endless. Lately, it has been developing rapidly in many of its distinct fields, especially in the field of biometric. the software developed for  this final project is just one fragment of an biometric application. it replicates the auditory system of a human being to indentify a speaker based on his/her voice, which is used as the input to the software. This application was developed using LPC (Linear Predictive Coding), also supported by DTW (Dynamic Time Warping) method. LPC  converts a sound signal so that needs  many step in the speaker identification Pre-Emphasize Filter, Frame Blocking, Frame Windowing,Auto Corellation Analysis, LPC Analysis dan Cepstral Coeffisient into several data vector which provide useful information in the speaker identification process. therefore, the support the Dynamic Time Warping method for speaker identification process. the software uses the Object Oriented Programming (OOP) concept and programmed in Borland Delphi 7.0. The application was tested on 10 people, each person was told to pronounce a set of 3 times predetermined words correctly, the test result showed that this application has  accuracy rate of 65% to 99% in speaker identification. the parameter values used both LPC and Dynamic Time Warping methods highly contribute to affecting the aplication’s accuracy rate.