Automatic Speaker Recognition System For Forensic Applications

Shaiful Adli, Yaakob (2013) Automatic Speaker Recognition System For Forensic Applications. Project Report. UTeM, Melaka, Malaysia. (Submitted)

	PDF (24 Pages) Automatic_Speaker_Recognition_System_For_Forensic_Applications.pdf - Submitted Version Download (724kB)
	PDF (Full Text) Automatic_Speaker_Recognition_System_For_Forensic_Applications.pdf - Submitted Version Restricted to Registered users only Download (2MB)

Abstract

This report is focus on the application of the automatic speaker recognition system for forensic application and it’s called forensic automatic speaker recognition. Forensic recognition aims or applies at the use of individualization.Our voice contains various characterization or parameters that convey information such as emotion, gender, attitude, health and identity. The speaker recognition for this particular project deals with the subject of identifying a person based on their unique voiceprint present in their speech data. There is another important stage that happens before voice feature extraction which called the pre – processing, where it ensures the voice feature extraction contains accurate information that conveys the identity of the speaker. For this particular project, the Mel Frequency Cepstrum Coefficient (MFCC) feature is used to extract the information or the characterization of the speech signal for a text dependent speaker identification system. Vector Quantization- Linde, Buzo and Gray (VQ-LBG) is used to quantized a number of centroids by using this particular algorithm. The codebook of speaker is constituted by these centroids. To be clear, MFCC are calculated in training phase and on the training session. The speaker is identified by using the concept of minimum Euclidean distance of the MFCC of each speaker in training phase to the centroid of individual speaker in the testing phase. All of development of algorithm is performed by using Matlab.The results shows high recognition rate when MFCC is used.

Actions (login required)

View Item