|
Title:
|
Speaker Recognition |
|
Author:
|
Saira Joseph, (Guide); Ansu, Mathew; Prathyusha, Vijayan K; Uthara, Chandran |
|
Abstract:
|
Speaker recognition is a task of identifying people from their voices. The key concept used here is the Mel- Filter Cepstral Coefficient. The cepstrum is a common transform used to gain information from a person’s speech signal.
First we read a sound signal. The power spectrum of this sound signal is extracted. The sound signal is mapped into Mel- scale .The Mel- scale approximates the human auditory system’s response more closely than the linearly spaced frequency bands. Next we reduce its dimension by vector quantization method. VQ is a process of mapping vectors from a large vector space to a finite number of regions in that space. Then the sound signals are saved in the data base. The new incoming sound signal is processed in the similar way. The signal is compared with the sounds stored in the data base. Comparison is done by Euclidian distance method. The threshold is set. The sound signal which is closer to the incoming signal is selected and speaker is recognized. External speaker is shown as mismatch |
|
Description:
|
Department of Electronics and Communication(For Further Reference Contact Department Library) |
|
URI:
|
http://hdl.handle.net/10562/1166
|
|
Date:
|
2008-06 |