You are on page 1of 6

Introduction:

The entire process of speech recognition can be broadly


split into two subsequent phases.
The training phase

The testing phase.

Two Phases-Block diagram

Basic Structure of Speech


Recognition system

Vector quantization
Vector Quantization is a process of mapping vectors
from a large vector space to a finite number of regions
in that space. Each region is called a cluster and can be
represented by its center called a codeword. The
collection of all code words is called a codebook.

Feature Extraction
The input voice signal is captured using a microphone in
real-time. Most of the energy content of voice signal is
around ..3 kHz to 4 kHz. So the captured signal should be
sampled at a rate greater than 8 kHz following the
Nyqiust rule.
The speech signal is a short time Wide Sense Stationary
Process. So each word is divided into frames of 20ms for
further processing. Hamming Window is applied to these
frames to minimize sharp discontinuities.

You might also like