Forensic speaker profiling: the study of supra-segmental features of Gujarati dialects for text – independent speaker identification

Abstract

In forensics, personal identification and criminal profiling have proved to be one of the important tasks in which variant features have been evaluated for the information related to individuals. Speaker profiling, identification of individuals through voice, is one such aspect of profiling. A qualitative analysis of vowels in different Gujarati-speaking regions was done to study the accents of spoken Gujarati. Although Gujarati is the official language of Gujarat state, the accent, lexicon and prosody vary across geographical areas. In this paper, we have subjected four major accent regions: Standard Gujarati, Kathiawari, Carotari and Kutchi dialects, which were chosen from different parts of the Gujarat state for this study. Standard Gujarati is considered as the basic language due the close approximation to the modern Standard Gujarati. In total, 175 males and 175 females were selected from each dialectal region keeping the selection criteria in mind, giving a total of 1400 speakers. The study was conducted in the context of text independent recognition of Gujarati language speech in four dialects. Qualitative analysis of Gujarati vowels, acoustic features, intonation and tone of a speaker as compared in this study could have the potential for identification of dialect accent and to provide additional aid in forensic criminal profiling.