A study on formant frequencies under different conditions and its effects on forensic speaker identification

Abstract

Forensic speaker identification is the comparison of known speech samples with the unknown one for the purpose of legal requirement. Various studies have been carried out on the acoustic parameters extracted from the voice signals however these studies were carried out in very controlled manner with limited samples due to the dynamic nature of voice. In this paper, the acoustic features of recorded voice signals of eleven speakers of having similar dialects in both normal reading and natural speech modes were studied using Computerised Speech Lab (CSL) on the basis of different vowel environment. Verbal probability numerical Scaling was worked out on the basis of matching points by using radar diagram pattern and analysed to find out its application in speaker identification. The above method will be simple and firsthand method to make the opinion for the purpose of forensic speaker identification and other forensic identification methods involving more than one parameter.