Abstracts Submitted: 510
Number of Users: 710
View Abstracts Submitted
Back to home Page
SPEAKER IDENTIFICATION USING TULU NASAL CONTINUANTS IN LIVE RECORDING AND MOBILE NETWORK RECORDING Arjun. M. S* and Usha. M** *Student, PG Diploma in Forensic Anthropology & Odontology, Yenepoya Medical College, Mangaluru - 18 **Junior Research Fellow, Department of Audiology & Speech-Language Pathology, , Kasturba Medical College, Mangaluru - 575001 Karnataka State, India email@example.com Abstract The most natural way to communicate is through speech. Recognizing individuals based on their voice is a mutual phenomenon. The voice of an individual can be recorded while planning, committing or confessing to a crime. It can be used to directly lay the blame on the suspect in the act of committing the crime. Forensic speaker identification or forensic speaker verification is an important terminology given to the legal procedure by which one recognizes if two or more recordings of speech are from the same speaker. The need to establish the identity for recognizing an individual from his or her voice is significant because of the legal consequences and forensic connections. The aim of the present study will be to obtain the correct percentage of speaker identification using Mel Frequency Cepstral Coefficients (MFCCs) on nasal continuants in Tulu speaking individuals using semi-automatic method. Tulu is a Dravidian Language spoken mainly in the south west part of the Indian state of Karnataka and also in the Kasaragod district of Kerala. The participants for the study will be twenty Tulu speaking neuro-typical adult males in the age range of 20-25 years. The material used will be meaningful Tulu words containing nasal continuants /m/ (bilabial) and /n/ (alveolar). The participants will be instructed to read the material five times each under two conditions (a) Live recording and (b) Mobile network recording. Each nasal will be subjected for MFCCs using semi-automatic speaker recognition software. The present study will be associated with semi-automatic speaker identification (SAUSI) where the investigator will select unknown and known (similar phonemes, syllables, words and phrase) speech samples, which has to be compared. Here the software will process nominated samples, extracts parameters and analyse them according to a particular logarithm program. The scrutiny will be made by the examiner, especially to elect whether samples are good enough or affected by factor such as noise, co-articulation, rate of speech etc. Thus the researcher will select comparable parts of speech samples for computerized acoustical analysis and will estimate the outcomes that the computer provides. This is typically faced in any forensic situation where an interaction of computer and investigator takes place. In SAUSI, software loaded to a computer automatically will extract comparable data from forensic samples after they have been selected by the investigator. The study will be compared under two conditions: (I) Live vs live recording and (II) Mobile network vs mobile network recording to check the correct speaker identification scores. The results of the present study would indicate relatively high percent of correct speaker identification using MFCCs in condition I and condition II. The obtained results would serve as potential measure in the forensic scenario for Speaker Identification using nasal continuants in Tulu Language.
© Copyright 2017 All Rights Reserved