Automatic Speech Recognition and Translation for Low Resource Languages

9Forensic Voice Comparison Approaches for Low-Resource Languages

Kruthika S.G.*, Trisiladevi C. Nagavi and P. Mahesha

Department of Computer Science & Engineering, S.J. College of Engineering, JSS Science & Technology University, Mysore, Karnataka, India

Abstract

In an ever-evolving technological landscape, there are lot of security attacks and frauds that are carried out online, thus, the prevalence of cybercrime has necessitated the use of digital forensics to analyze digital evidence spanning various areas such as image, speech, audio, DNA fingerprint, palm print, foot print, and network forensics. In the realm of forensic speaker recognition, forensic voice comparison (FVC) is a powerful tool that scrutinizes speech patterns in recordings of unknown criminals and known suspects. This chapter delves into the intricacies of traditional voice comparison approaches such as auditory, acoustic-phonetic, linguistic, and spectrographic analyzes for low-resource language (LRL). Traditional approaches use manual process and they consume a lot of time. Hence, in modern practice, semi-automatic and automatic approaches have been developed to improve the accuracy and efficiency of FVC. These approaches are based on time and frequency domain model and utilize deep learning techniques to identify suspects. This chapter provides a comprehensive overview of traditional and modern techniques used along with challenges, motivation, datasets, applications, and research scope for LRL.

Keywords ...