Home // International Journal On Advances in Systems and Measurements, volume 10, numbers 3 and 4, 2017 // View article

A Practical Forensic Method for Enhancing Speech Signals Drowned in Loud Music

Authors:
Robert-Alexandru Dobre
Radu-Mihnea Udrea
Cristian Negrescu
Dumitru Stanomir

Keywords: adaptive algorithms; system identification; noise reduction; multimedia forensic

Abstract:
Recording audio or video is nowadays easier than ever. Almost every phone can do this task with high quality. This has some serious implications in forensic: almost every dialogue or event can be recorded and used as evidence in trials. The problem is that editing multimedia content has also become a very accessible operation. The advances of editing software make it possible with very convincing results for the untrained audience. Forged recordings could be used in trials. The need for multimedia forensic is imminent. There are two main directions of this field: probe authentication and noise reduction. This paper presents the research activities conducted to extract speech signal masked by loud music. The developed system is based on an adaptive system identification configuration. Various scenarios are studied showing the advantages and disadvantages of the adaptive algorithms that were tested. The influence of the acoustic environment over the performances of the proposed system is also studied and the results can help to determine if placing a microphone in a specific room could be used to intercept a speech.

Pages: 194 to 205

Copyright: Copyright (c) to authors, 2017. Used with permission.

Publication date: December 31, 2017

Published in: journal

ISSN: 1942-261x