by Khai Tran
Research into the concept of Speech technology began as early as 1936 at Bell labs.
But at that time, Bell labs... more
Research into the concept of Speech technology began as early as 1936 at Bell labs.
But at that time, Bell labs abandoned the project because of taking wrong path: artificial
intelligence is the key to the success. There was no further progress until 1970s;when
Lenny Baum of Princeton University invented Hidden Markov Model (HMM) which
provided a statistical-based approach to generate text from speech. From that point till
now, HMM, with the help of remarkable development in terms of computer processing
powers, has been widely used by companies and organizations to develop their own
speech recognition systems.
Currently, USA is the leading country in Automatic Speech Recognition (ASR) field of
research. Their ASR systems are known to be capable of recognizing natural speech in
ideal conditions with accuracy above 90%. There are as well other ASR systems have
been developed for other languages, for example Mexican, Australian, Spanish, French
…. However, because of language particularity and matters of business competition,
each country or company have their own recognition technology and normally they keep
it confidential within the country or organization.
In Viet Nam, ASR is quite new and there are research groups all over country trying to
develop ASR for Vietnamese. However, they are working independently with their own
engines and not yet to release any practical product. Therefore, with the purpose of
contributing to the research community, we are trying to develop an open and practical
Vietnamese Speech Recognition Technology (VSRT).
The purpose of the thesis is to propose a practical method of recognizing Vietnamese
words and put it into real software application that will be able to recognize digits (0-10).
The two first chapters of the thesis serves as a preparatory document with the purpose
of gathering, classifying and analyzing various sources of knowledge about ASR
technology and Vietnamese language necessary to build a new Vietnamese digit
Recognition system. Proposal, technical details and evaluating results will be appeared
in chapter 3 and 4 of the thesis.
Download (.pdf) 
