Exploiting Speaker Recognition for Home Security Applications

Abstract

Nowadays privacy is considered as an important issue for all individuals Conventional approach of utilizing the combination of username and password for authentication is becoming less secure On the other hand biometrics become new features for authentication because biometrics are inherently unique and measurable characteristics that can be used to identify a person Among several types of biometrics it is generally easier to get voice data of a person In order to improve home security in our research using speaker recognition technology the technology is to identify a person utilizing the characteristics of human voice Speaker recognition techniques can be effectively to extraction the person’s vocal tract features Their vocal tract shapes larynx sizes and other parts of their voice production organs are different no two individuals sound identical In prior works the Mel-Frequency Cepstrum Coefficients can describe the vocal tract characteristics and easy to captures vocal tract characteristics more effectively The extracted features by vector quantization approach to create voice modeling During the identification a speech sample or utterance is compared against a previously created voice model Our prototype system has two main functions including speaker recognition and history review Experimental results show that our system can instantly and accurately identify family members in the home environment Moreover strangers can be detected so as to actively alert family members

Date of Award	2016 Aug 27
Original language	English
Supervisor	Wei-Guang Teng (Supervisor)