Automatic Speech Recognition (ASR), which is concerned
with the problem of recognition of human speech by a machine, is the core
of a natural man-machine interface. In spite of this realization, the
progress in this field has not been smooth due to the inherent complexity
and difficulty of the problem. Environmental noise and language structures
are two important issues which highly affects the recognition performance.
Speech signal processing is the first step in developing ASR. Second step
is to model those features for recognition. Broadly speaking there are
three approaches used for ASR, namely:
The acoustic-phonetic approach
The pattern recognition approach
The artificial intelligence approach
In these approaches, statistical pattern recognition approaches are evolved
as best . Some of them are VQ (vector Quantization) and HMM (Hidden
Markov Models) .
A speaker dependent voice recognition system has been developed
for drawing and manipulating geometrical figures using continuous
speech. The system has the provision for accepting a new user and
training his/her voice.