Likelihood Ratio Estimation and Validation for Discrimination of Ten Hindi Vowels in Forensic Speaker Identification Jamwal Aarti1, Sehrawat Surender Singh2,*, Kaur Gurleen3, Verma Rajesh4 1M.Sc. Forensic Science and Toxicology, Department of Forensic Science, Chandigarh University, Mohali, Punjab 2PhD Scholar, Department of Cardio-thoracic and Vascular Surgery, PGIMER, Chandigarh 3Assistant Professor, Department of Forensic Science, RIMT University, Mandi Gobindgarh, Punjab 4Deputy Director, Regional Forensic Science Laboratory, Mandi, Himachal Pradesh *Corresponding Author, Dr. Surender Singh Sehrawat, PhD Scholar, Department of Cardio-thoracic and Vascular Surgery, PGIMER, Chandigarh, Email: surendersehrawat6@gmail.com
Online Published on 12 January, 2024. Abstract The forensic voice comparison process is undergoing a paradigm shift in terms of the evaluation of the evidence. The traditional “match” or “no match” approach is being replaced by the likelihood ratio based approach. The new approach has been incorporated in the best practices manuals of many European & American forensic organizations. But this approach has not seen much development in many of the Asian Countries including India. Technical speaker recognition now involves the acoustics-phonetic approach (semi-automatic) and the automatic approach. The present study has been undertaken to assess the variability of the Hindi vowels that helps in the text independent voice comparison. Top Keywords Forensic, Speaker identification, Vowels, Log likelihood ratio cost approach, Formants. Top |