The following information was submitted:
Transactions: WSEAS TRANSACTIONS ON SIGNAL PROCESSING
Transactions ID Number: 52-525
Full Name: Georges Jabbour
Position: Ph.D. Candidate
Age: ON
Sex: Male
Address: Av. Alberto Carnevalli, Nucleo Pedro Rincon Gutierrez, Universidad de Los Andes, Facultad de Ingenieria. Merida
Country: VENEZUELA
Tel: 0582742402989
Tel prefix:
Fax:
E-mail address: jabbour@ula.ve
Other E-mails: gjabbour50@hotmail.com
Title of the Paper: A hybrid system based on Hidden Markov Models and Support Vector Machines with forward learning for phone recognition in Venezuelan continuous speech
Authors as they appear in the Paper: Georges Jabbour, Luciano Maldonado and Maria Sarmiento
Email addresses of all the authors: jabbour@ula.ve, maldonaj@ula.ve, marialaurasc@gmail.com
Number of paper pages: 11
Abstract: The performance of an automatic speech recognizer based on Hidden Markov Models (HMMs) and Support Vector Machines (SVMs) is here compared to the performance of two other recognizers: one based on the HMMs only (the HMM recognizer), and the other is a hybrid model based on HMMs and SVMs (the HMM/SVM recognizer), the architecture and training of which differ from those of the recognizer proposed in this paper. The recognizer we propose here is called the SVM/HMM Hybrid Model with forward learning (the SVM/HMMFL recognizer). The three recognizers were programmed using Matlab and trained with Venezuelan continuous speech, the speech signals of which form part of the SpeechDat European project. The phone was chosen as the acoustic training unit. The recognition tests performed showed a significantly better performance of the hybrid recognizers when compared to that of the recognizers based on the HMMs only. Of the two hybrid recognizers, the best results were obtained !
with the SVM/HMMFL recognizer.
Keywords: Automatic Speech Recognition, Support Vector Machines, Hidden Markov Models, Hybrid SVM/HMM Model, Artificial Intelligence
EXTENSION of the file: .doc
Special (Invited) Session:
Organizer of the Session:
How Did you learn about congress: Automatic Speech Recognition, Signal Processing, Artificial Intelligence
IP ADDRESS: 190.76.56.141
