Model Fonem dengan Pendekatan Distribusi Normal untuk Pengenalan Kata Menggunakan MFCC sebagai Ekstraksi Ciri

Abstract

Voice recognition is a field study in voice processing. Research on voice signal has several methods of processing, one of them is the Mel Frequency Cepstrum Coefficients (MFCC). The purpose of MFCC is to extract voice features. The goal of this research is to apply MFCC as a feature extraction and the normal distribution as a method for word recognition. The first step of the research is data reprocessing, and then data extraction using MFCC. Afterwards, the normal distribution method is used to process the data. From the result, it can be concluded that the normal distribution method can be used for word recognition.The results obtained from the word recognition using normal distribution and MFCC as feature extraction have the highest accuracy of 100% for the phonemes /g/, /n/, /p/, /v/, /x/, /y/ and the lowest accuracy of 67% for the phonemes /a/ and /k