Speech Recognition Using MFCC and VQLBG

Abstract

Speaker  Recognition  is  the  computing  task  of confirmatory a user’s claimed  identity mistreatment characteristics extracted  from  their  voices.  This  technique  is  one  of  the  most helpful  and in style  biometric  recognition  techniques  in  the  world particularly connected  to  areas  in that security could be a major concern. It are often used for authentication, police work, rhetorical speaker recognition and variety of connected activities. The method of Speaker recognition consists of two modules particularly feature extraction and have matching. Feature extraction is that the method during which we have a tendency to extract a tiny low quantity of knowledge from  the  voice  signal  that will  later  be  used  to  represent every  speaker.    Feature  matching involves  identification  of  the  unknown  speaker  by scrutiny  the  extracted options  from his/her voice input with those from a collection of identified speakers. Our projected  work  consists  of  truncating  a  recorded  voice  signal,  framing  it,  passing  it through  a  window perform, conniving  the  Short  Term  FFT,  extracting  its options  and Matching it with a hold on guide.  Cepstral constant  Calculation  and  Mel  frequency Cepstral  Coefficients  (MFCC) area unit  applied  for  feature  extraction  purpose.VQLBG (Vector Quantization via Linde-Buzo-Gray) algorithmic rule is used for generating guide and feature matching purpose.