Intra and inter speaker variation of acoustical parameters in text dependent and text independent forensic speaker identification on the basis of radar diagram and verbal probability numerical scale

There is large scale increase in application of Biometric system in various fields like security, defence, payroll, forensic etc., and now a day the voice biometric system plays crucial role in the field of forensic identification. However, it is very tough to make definite opinion in Forensic Speaker Identification (FSI) due to the dynamic nature of human voice signal. Hence various tools and methods have been applied to study the variations in the acoustic, phonetic parameters and make opinion on whether the speakers are same or otherwise. Most of tools and methods are helpful in laboratory conditions but on forensic conditions the process of identification becomes more complex due to legal implications. We applied new tool called “radar diagram to analyse the acoustical data of speech samples recorded from 11 persons and verbal probability numerical scale to make opinion on speaker identification. Intra and inters-speaker variations are discussed in text-dependent and text-independent methods; two different forensic scenario. The method adopted was found useful in terms of its simplicity and effectiveness while making opinion during FSI.

Here we applied these tools to analyse and make opinion on the formant frequency of six prominent vowels extracted from the speech samples of 11 persons. In this study the radar diagram was used in both the text dependent method and text independent speaker identification. Further this method was applied on both intra and inter-speaker variations.