Study on Bangla Phonetic Features

Abstract

Speech recognition is one of the most important computer technologies for Automatic Speech Recognition (ASR) model. ASR model have to enables device for recognize and very better known as Bangla word. ASR model check sound and digitizing and also matching the word pattern for stored patterns. We have to see after recognize normal speech not better than discrete speech. Now, we can detect continuous speech (Normal Speech) for used a system. Its helps people about working for disabilities. We used a system which is new approach for buildup in top position Bangla ASR model. Its MFCC based system used for better word recognition performance. Maximum cases used in small no. of speaker, but Bangladesh most of the wide area are used in native language. We can experiments about this language as a inputs are Mel-Frequency Cepstral Coefficients (MFCCs) and result will be based on Hidden Markov Model (HMM) classifier for word recognition performance. We can experiments about other ASR models, firstly we used MFCC-39 based classifier and now we can used Articulatory features AF-22 based classifier for male and female voice. Then, we can see word correct rate and accuracy will be much better than MFCC- 39 based classifier. That’s why our suggest system help for detect gender voice and independent fact.