Pendeteksian Kata dengan MFCC sebagai Ekstraksi Ciri dan Codebook sebagai Pengenalan Pola

Abstract

Word detection is the process of identifying words spoken by someone. It is useful for communication between computers and humans. MFCC technique was used to extract features from voice signal and compare it to the sound signal in the database. Codebook is the process of producing small characteristic vectors was used to match unknown words to words in the database. The purpose of this research was to quantify detection accuracy of spoken words and compare 16 codewords and 32 codewords. The experiment produced an average detection accuracy of 98%. We conclude that detection using MFCC and codebook can achieve good performance. And the performance of the 16 codewords is better than 32 codewords for word detection.