TY  - THES
T1  - Automatic isolated word recognition system and phoneme-level segmentation of the Filipino speech corpus using the hidden Markov model toolkit
A1  - Chua, Ivy Jhoanna A.
A2  - Eustaquio, Nino D.J
LA  - English
YR  - 2006
UL  - https://tuklas.up.edu.ph/Record/UP-99796217603116287
AB  - Automatic speech recognition systems (ASR) are widespread in today's modern world and have proven its use in many applications. A prerequisite in the development of such systems are segmented and transcribed speech corpus which, when done manually, is a tedious work. Hence, the objective of this project is to develop a system that will automatically recognize speech signals at the word-level, and segment the words of the Filipino Speech Corpus (FSC) word sub-corpus into phoneme-level. The Hidden Markov Model Toolkit (HTK) developed at Cambridge University Engineering Department (CUED) will be used for this purpose. To create HMMs, Mel Frequency Cepstral Coefficients (MFCC) is extracted using the tools provided by HTK, and in turn will be used as input feature vectors. The system will be trained using these feature vectors to recognize words in word level. From the results of the word-level recognition, a phoneme-level recognition system will be developed which will be used for the phoneme segmentation process. The output of this project is an acceptable word recognition rate of the ASR system and the phoneme-level segmentation of the FSC word sub-corpus.
CN  - LG 993.5 2006 E64 C48
KW  - Word recognition.
KW  - Phonemics.
KW  - Automatic speech recognition.
KW  - Markov processes.
KW  - Word recognition system.
KW  - Filipino speech corpus.
KW  - Markov Model.
ER  -