Voice Comparison Approaches for Forensic Application: A Review

Abstract

Advancement in technology has led to various types of online security attacks and frauds. Evidences produced is in digital form and the field of analysis is called digital forensics. It focuses on identifying, acquiring and analyzing electronic evidence. However, there are many research works and techniques under the speech forensics domain such as Forensic Speaker Recognition (FSR) and Forensic Voice Comparison (FVC) for supporting the crime investigation. Within the broader domain of speech forensics, the paper focuses specifically on FVC, which adopts a trace sample to examine speech patterns in a recording of a known suspect voice. This article provides an overview of generic approach, traditional and semi-automated methods for FVC, including auditory, acoustic, phonetic, linguistic and spectrographic methods. In present scenario there are many automatic approach proposed using deep learning to identify the criminals using time and frequency domain methods. They present the efficient results and there is a scope for improvement with reference to FVC. The research also examines various challenges, applications and deliberations. Finally, the paper concludes with discussions and future directions.