Speech Recognition Fundamentals
Speech
is a very natural and an efficient manner of communication for humans. The
study of speech recognition concerns itself with the development of systems
that emulate the human’s natural ability to process such an exchange of
information. Researchers have found that this is not a simple task. In the
early 1970s, commercial systems were developed to recognize isolated words.
These were small systems, which employed a vocabulary of 10 to 100 words.
Towards the late 1980s, researchers at IBM developed a system capable of
recognizing continuous (naturally spoken) and isolated utterances from a
vocabulary of 20,000 and 50,000 words respectively.
Our system works on statistical modeling of isolated words, but lacks in
grammatical modeling hence can only be used in Command Mode Recognition.
Though the system incorporates Dictation Mode Recognition, it is not useful
unless the system has Grammar Model to type correct sentences uttered by the
user.
Speech
Trainer Engine
Speech Trainer Engine
intends to train the system for a particular user. The system needs to be
trained for as many of words that user wishes to recognize.
Speech Recognition Engine
Speech Recognition Engine intends to type a word uttered by the user. One
has to make sure that he trains the system for the intended words to be
recognized.
Install
Section
Click here to download the
engine.
|
