AES Technical Committee

Audio Forensics

Chair:    Jeff M. Smith    (Send Email)
Vice Chair:    Daniel Rappaport    (Send Email)
Vice Chair:    Eddy Bogh Brixen    (Send Email)
Forum:    Committee Discussion Forum

Current Areas Of Work

  • To extend knowledge of audio forensic practice and techniques world-wide
  • To identify areas requiring additional research
  • To bring together professionals and students in the area of audio forensics
  • To encourage publication of papers, and to arrange tutorials, conferences and workshops, within the AES
  • To support standards committees on forensic audio via fostering of refereed journal publications

Recent/Planned Activities

Another successful conference of the AES on Audio Forensics was held June 12-14 in London. See the website for specific program and sponsor details: Proceedings from this conference, as well as the previous audio forensics conferences (26, 33, 39, 46) are available here:

At the AES 137th Convention in Los Angeles, a Tutorial on "Audio Forensics: An Overview". ABSTRACT: This tutorial featured several presenters engaged in various areas of audio forensics in lively discussion geared toward experts learning from one another and to benefit the introductory attendee. Tutorial Chair, Jeff M. Smith (National Center for Media Forensics, CU Denver) and chair of the Technical Committee on Audio Forensics, presented on Speaker Analysis and the application of Bayesian likelihood. Catalin Grigoras (NCMF, CU Denver) presented on the best practices and future challenges in forensic audio authentication. Gordon Reid (CEDAR Audio Ltd.) presented on noise reduction and speech enhancement techniques. Eddy Brixen (EBB consult) presented on auditory crime scene analysis. Finally, Durand Begault (Audio Forensic Center) wrapped up some loose ends with a discussion of forensic audio miscellany including musicological forensics, warning signal audibility, acoustics, and more.

Mission Statement

Audio Forensics refers to engineering and scientific analysis, evaluation and presentation of audio and acoustic evidence in a judicial inquiry normally leading towards a presentation in court. The complex challenges faced by the field transcend borders and jurisdictions, and becomes increasingly complex with the proliferation of new types of digital media. In order to establish a reliable scientific basis for the procedures and analysis of the outcomes of audio forensic analysis, the technical committee meets at AES conventions to plan conferences, workshops and tutorials; to encourage dissemination of scientific information through peer-reviewed journal and conference publications and posters; and to communicate educational opportunities in the field. Presently, the technical committee meetings for audio forensics are open to all attendees.

Emerging Trends in Audio Forensics

Last updated: 10/31/2011

Enhancement -

The enhancement of forensic audio recordings remains the most common task for forensic audio practitioners. The goal of forensic audio enhancement is to increase intelligibility of voice information or improve the signal to noise ratio of a target signal by reducing the effects or interferences that mask it. Many tools are available through various software developers with the most common being noise reduction- either adaptive or linear. Difficulties in this area are caused by lossy data compression common to small digital recorders, data compression and bandwidth limited signals in telecommunications, and non-ideal recording environments common to surveillance and security. One growing area of research is the assessment of speech intelligibility with multiple papers presented on the topic at the AES 39th Conference on Audio Forensics in 2010.

Authentication -

The majority of audio media presented to the forensic examiner are digital recordings on optical disc, HDD, flash memory, and solid-state recorders. However, the analysis of analog tape cassettes and microcassettes is still required of examiners. In the area of forensic media authentication, digitally recorded audio files may be subject to various kinds of manipulation that are harder to detect than those in the analog domain. This leaves the forensic audio examiner with new challenges regarding the authentication of these recordings. Many new techniques have been developed in recent years for use in these analyses. These techniques continue to be published and presented through the AES Journal and proceedings of AES Conferences and Conventions. Among these techniques is the analysis of the Electric Network Frequency component (ENF) of a recording. If present, the remains of the ENF may be compared to a database of ENF from the same grid to authenticate the date and time the recording was made. In addition to automatic database comparison, it is possible to learn several other things from ENF analysis including whether portions of the recording were removed, if an audio recording was digitized multiple times, how the recorder was powered, and more.

Recent developments in digital audio authentication also include the Compression Level Analysis of an audio recording to determine if an uncompressed file had been previously subject to data compression or if the compression level present is consistent with an authentic recording. Also, a technique for determining the presence of butt-splice edits has been presented. In the digital domain, as in the analog, auditory and spectral acoustic analysis continues to be necessary. However, it is also clear that analysis of the digital data that makes up a recorded audio file including its header and file structure must be exploited to ascertain a digital recording’s authenticity.

Speech and Speaker Analysis -

The analysis of speech and speakers present on audio recordings is a large domain that intersects many industries including forensics and security. The analysis of speakers present in recordings to ascertain identity continues to be a common request of forensic audio examiners. However, “identifying” persons in a 1:1 comparison is not supported within the scientific community that favors “recognition” of persons based on extracted features relative to a background model representing a population of speakers. Automatic systems based on cepstral coefficients, Gaussian Mixture Modeling, and likelihood ratios employ robust and validated techniques for speaker recognition. This quantitative approach better measures and takes into account intra- and inter-speaker variation. When used in a forensic environment where trained examiners base conclusions on likelihood ratios, this technique is valued greatly over other qualitative analyses.

The capability of a system to process multitudes of audio signals and sort them based on language, topic, speakers present, and acoustic environment continues to progress with many new advances. An interesting area of research and its application in audio forensics is Computational Auditory Scene Analysis (CASA). This field of audio processing is interested in developing machine systems that perform automatic signal separation using principles derived from the human auditory system's perceptual abilities to understand audio scenes. CASA systems have already proven very useful as pre-processors for automatic speech recognition systems and in hearing aids. New areas of study include their use in audio forensics. Also, automatic speaker segmentation based on extracted spectral features and statistical modeling can help automated systems tasked with speech and speaker recognition.

Other Considerations -

Since the fundamental aspect of forensic audio is its application to law with the litigation process benefitting from audio enhancement and analysis, it is important for the practitioner working with forensic audio to be aware of this process and the need for proper evidence handling and laboratory procedures. As digital audio proliferates so to have the identification of proper practices for imaging media, hashing file duplicates, and recovering and/or repairing corrupt or carved files.

Additionally, it is not only common for forensic audio to be played in a courtroom but for typed transcripts of recorded conversations to be prepared for the individuals involved in a case; the lawyers, judge(s), and/or jury. Specific to these needs, there are developments in addressing the inherent bias present in the human preparation of these transcripts. Also, the forensic audio practitioner must be aware of the audio samples being presented taking into consideration courtroom acoustics, psychoacoustics, and the hearing abilities of these individuals.

AES Activities -

Numerous papers on audio forensics appear in the Journal of AES and are presented at AES Conventions each year. Additionally, there have been three AES Conferences on audio forensics since 2005(AES 26th, 33rd, and 39th) and the next will be in Denver, CO in 2012. Additionally, regular workshops and tutorials appear at AES Conventions. At the AES 130th Convention in London there was a tutorial on forensic audio enhancement. At the AES 131st Convention in New York there was a workshop on forensic audio authentication.

Meeting Report:

These documents do not necessarily express the official position of the AES on the issues discussed at these meetings, and only represent the views of committee members participating in the discussion. Any unauthorized use of these publications is prohibited. Authorization must be obtained from the Executive Director of the AES: Email, Tel: +1 212 661 8528, Address: 551 Fifth Ave., Suite 1225, New York, New York 10176, USA.

2014-10-16     137th AES Convention TC-AF Meeting Report
Description: Minutes from TC-AF meeting in LA on Oct. 11, 2014

2014-5-23     135th AES Convention TC-AF Meeting Report
Description: TC-AF meeting minutes from the 135th AES Convention, New York, Oct. 2013

2013-9-3     134th Convention TC-AF Meeting Report
Description: May 2013 TC-AF meeting minutes

2013-9-3     133rd Convention TC-AF Meeting Report
Description: October 2012 TC-AF meeting summary

2013-9-3     132nd Convention TC-AF Meeting Report
Description: April 2012 meeting minutes

2013-9-3     131st Convention TC-AF Meeting Report
Description: October 2011 TC-AF meeting minutes

2013-9-3     130th Convention TC-AF Meeting Report
Description: May 2011 meeting minutes

2013-9-3     129th Convention TC-AF Meeting Report
Description: November 2010 TC-AF meeting minutes

2013-9-3     128th Convention TC-AF Meeting Report
Description: May 2010 TC-AF meeting minutes

2013-9-3     127th AES Convention TC-AF Meeting Report
Description: October 2009 TC-AF meeting minutes

2013-9-3     125th AES Convention TC-AF Meeting Report
Description: Oct. 2008 TC-AF meeting minutes

Committee Members

 Christoph Musialik  Durand Begault  Gordon Reid 
 Rob Maher  W.A. Deutsch  Douglas S. Lacey 
 Philip Esterhazy  Eddy Bogh Brixen  Anibal Ferreira 
 Lise-Lotte W Tjellesen  J. Keith McElveen  Wes Dooley 
 Wanda Newman  Robert-Eric Gaskell  Margaret Luthar 
 Myke Roy  Erik Achrén  Ronny Max 
 Tom Owen  Mark Stebbeds  Robert Lusk 
 Ryan Hansen  Davis Appelt  Todd Makinen 
 Lars Clausen  Pascal Lorenz  Kuldip Raj 
 Thor Legvold  Florian Raschke  Sebastian Schneider 
 Michaela Schuster  Jeff M. Smith  Phil Manchester 
 Truls Birkeland  Jacob Setinhour  Andrey Barinov 
 Hafiz Malik  Matthew Owens  Matthew Eardley 
 Norm Broner  Neil Muncy  Kristin Bidwell 
 Doug Gregory  Catalin Grigoras  Felix Klemke 
 Dan Nyberg  Jeff Harris  Vlado Delic 
 Daniel Rappaport  Federico Jankilevich  David Hallimore 
 Steve Saunders  Thomas Guzman-Sanchez  Bruce Koenig 
 Marisa Dery  Alexander Zaknewski 

To request membership in this Technical Committee please email the Chair by using the link above.

Facebook   Twitter   LinkedIn   Google+   YouTube   RSS News Feeds  
AES - Audio Engineering Society