Abstract: Speaker Recognition System is increasingly being deployed as a more natural means for recognition of people. Speaker Recognition (SR) is the process of automatically recognizing the person speaking on the basis of the information obtained from the speech features. In today world scenario, speaker recognition system is very popular in voice verification for identity and access control to services. Speaker identification and verification are basic part of speaker recognition system. Basically it is a process of automatically recognizing the identity of speaker on the basis of individual information extracted from speech signals. Our main purpose is to reduce error in the system and increase the efficiency. In this report, it is done with the help of different feature extraction algorithm. We have introduced a new approach for speaker recognition. This new system firstly frames the speech signals and then these signals are compressed using DWT for noise reduction and better sampling frequency. Furthermore, features of compressed signals are extracted with the help of Discrete Wavelet transform (DWT), Non-Linear Energy operator (NEO) based DWT, and Irregular Discrete Wavelet Packet transform (IRR DPWT). The distance metrics incorporated are Delta Bayesian Information criteria (delta BIC). At the end results are evaluated with Detection Error Tradeoff (DET) curve and Receiver Operator Characteristics (ROC) curve by finding the area under curve (AUC). The main purpose of speaker identification system is to provide a reliable decision, accept or reject a speaker, given a claimed identity and a recording of a speaker phrase. After comparing area under curve we can say that the best result is shown by DWT NEO with BIC distance metric.
Keywords: Speaker Recognition System, Discrete Wavelet transform (DWT), BIC distance metric.
Title: Speaker Identification using Discrete Wavelet Packet Transform Technique with Irregular Decomposition and BIC
Author: Jyoti Devi, Sukhvinder Kaur
International Journal of Electrical and Electronics Research
ISSN 2348-6988 (online)
Research Publish Journals
