2007 May, Volume 55 Number 5

CONTENTS

PAPERS

Background, Concept, and Architecture for the Recent MPEG Surround Standard on Multichannel Audio Compression
Jeroen Breebaart, Gerard Hotho, Jeroen Koppens, Erik Schuijers, Werner Oomen, and Steven van de Par   331
A description, analysis, and subjective evaluation of the new MPEG Surround standard show its flexibility and utility. A multichannel signal is transformed into a conventional stereo pair with an additional side channel containing spatial information, thereby being backward compatible with legacy techniques. Subjective listening tests showed that this technique produces superior perceptual quality compared to conventional techniques. The required bit rates for multichannel audio are comparable to those employed by conventional coders for stereo content. Moreover, by transmitting spatial information parametrically, the system provides for spatial-format conversion.

Forensic Enhancement of Digital Audio Recordings
Bruce E. Koenig, Douglas S. Lacey, and Steven A. Killion   352
With the dramatic advancement of digital tools over the last two decades, forensic audio processing has acquired new methods and protocols. Audio recordings continue to be an important part of the judicial process. Unintelligible recordings made in less than ideal conditions can be made useful for legal proceedings. An ideal forensic facility should have trained examiners, appropriate laboratory space, modern digital filters, formal evidence-handling practices, and experts prepared to give testimony. Four examples demonstrate the applications of modern forensic audio processing.

Intelligent Preprocessing and Classification of Audio Signals
Mingsain R. Bai and Meng-Chun Chen   372
The ease of acquiring large quantities of audio files and the dramatic reduction in the cost of storage have created a need for audio search tools that can perform a browsing function. Automatic tagging of audio would be useful. Such systems are based on feature extraction and algorithmic classification. The proposed system first extracts nineteen features based on spectral, temporal, and statistical attributes of the signal. These features are then used as input to any of the following: nearest neighbor rule, artificial neural networks, fuzzy neural networks, and hidden Markov models. The system also offers optional preprocessing functions for blind source separation, vocal removal, and denoising. Empirical tests results demonstrate the performance of this approach.

ENGINEERING REPORTS

Estimating the Instantaneous Frequency of Sinusoidal Components Using Phase-Based Methods
Mathieu Lagrange and Sylvain Marchand   385
A robust method to estimate the short-term spectra of an audio signal is important in many applications, such as sinusoidal modeling of music. Using a phase-based estimation approach, the authors prove five of these techniques are actually theoretically equivalent. But in a practical application, differences in performance among various estimation approaches result from assumptions about signal complexity and the implementation of the algorithms. A Hilbert filter in a preprocessing stage can improve the precision of frequency estimation.

STANDARDS AND INFORMATION DOCUMENTS

AES Standards Committee News   400
Audio metadata; loudspeaker measurement; audio connectors

FEATURES

Wireless Microphones in Live Sound Applications   402

DEPARTMENTS

News of the Sections   409
Sound Track   416
New Products and Developments   418
Upcoming Meetings   418
Available Literature   419
Membership Information   421
Advertiser Internet Directory   423
In Memoriam   430
Sections Contacts Directory   432
AES Conventions and Conferences   440

FPO For the cover:

2007 May, Volume 55 Number 5

spine: 2007 May, Volume 55 Number 5