Home | Technical Program | Exhibition | Visitors | Students | Press
Last Updated: 20060404, meiP18 - Posters: Signal Processing and High-Resolution Audio
Monday, May 22, 11:00 — 12:30
P18-1 Advanced Cataloging and Search Techniques in Audio Archiving—Helge Blohmer, VCS Aktiengesellschaft - Bochum, Germany
Ever since the processing capabilities of computers reached the point where audio indexing and searching became possible using techniques beyond simple, manually entered textual annotation in the late 1980s, researchers have been developing such methods with varying degrees of success. Yet even today, the actual workflow in audio archives is dominated by text entry for cataloging and keywords for searching with few or none of the new methods having achieved any practical relevance. This paper evaluates a number of techniques, both those that enhance textual retrieval and those that seek to supplant it, toward their suitability for real-world audio archiving tasks with special focus on their suitability for a short-term implementation and seamless integration into existing archive workflows.
[Poster Presentation Associated with Paper Presentation P13-1]
Convention Paper 6726 (Purchase now)
P18-2 MP3 Window-Switching Pattern Preliminary Analysis for General Purposes Beat—Antonello D’Aguanno, Goffredo Haus, Giancarlo Vercellesi, Università degli Studi di Milano - Milan, Italy
This paper analyses the dependency of the window-switching pattern versus different encoders, bit rates, and encoder quality features. We propose a simple template-matching algorithm to solve beat tracking contest in music with drums. This algorithm uses windows-switching pattern information only. Commonly in a beat-tracking system the window-switching pattern is used to refine the results of a frequency evaluation. Furthermore, this paper wants to demonstrate the reliability of the window-switching pattern to solve beat-tracking problems in music with drums independently from encoders, bit rates, encoders’ quality features, and frequency analysis. This paper confirms the window-switching pattern is adequate information in a beat-tracking contest at every bit rate and for every encoder.
[Poster Presentation Associated with Paper Presentation P13-3]
Convention Paper 6728 (Purchase now)
P18-3 Application of MPEG-4 SLS in MMDBMSs—Requirements for and Evaluation of the Format—Maciej Suchomski, Klaus Meyer-Wegener, Florian Penzkofer, Friedrich-Alexander University Erlangen-Nuremberg - Erlangen, Germany
Specific requirements for audio storage in multimedia database management systems, where data independence of continuous data plays a key role, are described in this paper. Based on the desired characteristics of the internal format for natural audio considering especially long-time storage, where the storage must be lossless, allowing among others easy upgrade of the system, the new MPEG-4 scalable lossless audio coding (SLS) is briefly explained. It is then evaluated w.r.t. the discussed requirements, looking at characteristics and processing complexity of the algorithm. Some suggestions of the possible modifications are given at the end.
[Poster Presentation Associated with Paper Presentation P13-4]
Convention Paper 6729 (Purchase now)
P18-4 Applying EAI Technologies to Bimedial Broadcast Environments: Challenges, Chances, and Risks—Michael Zimmermann, VCS Aktiengesellschaft - Bochum, Germany
More and more broadcast companies try to optimize their production environments by enforcing bimedial workflows. The recent applications and tools on the other hand only have poor integration interfaces to achieve this goal. EAI, originally focusing on the integration of legacy systems, has become a mature toolset to integrate various systems and offering tools and applications to ease integration. This paper shows the possibilities and limits of EAI in bimedial broadcast environments.
[Poster Presentation Associated with Paper Presentation P13-5]
Convention Paper 6730 (Purchase now)
P18-5 Personal Audio Headrest—Chiho Chung, Samsung Electronics Co. Ltd. - Suwon, Korea; Steve Elliott, ISVR, University of Southampton - Southampton UK
Active noise control was implemented using loudspeakers embedded in the headrests of two adjacent seats. The goal of this project was to create a quiet zone surrounded by headrest 1, which was free from noise caused by the adjacent loudspeaker mounted in the next-to headrest 2. While headrest 1 was generating noise, headrest 2 was designed to cancel out noise that was being generated by headrest 1, by driving the anti-noise signal through the loudspeaker, without using earphones/headphones. Control source, the loudspeaker of headrest 1 generating anti-noise, was made from FIR convolution with the electrical signal going to the primary source, the loudspeaker of headset 2 generating target-noise. Implementing both primary and control sources results in a 20 to 30 dB noise reduction throughout the targeted frequency range (2 kHz and below) in terms of squared acoustic pressure.
Convention Paper 6768 (Purchase now)
P18-6 Accidental Wow Evaluation Based on Sinusoidal Modeling and Neural Nets Prediction—Przemyslaw Maziewski, Lukasz Litwic, Andrzej Czyzewski, Gdansk University of Technology - Gdansk, Poland
In this paper an algorithmic approach to the wow defect characteristic evaluation is presented. The approach is based on a sinusoidal analysis comprising both amplitude and phase spectra processing. The frequency trajectories depicting the distortion are built on a basis of amplitude, frequency, and phase dependencies and are further used for wow characteristic evaluation. Additionally the experiments concerning the neural-network-based prediction applied to the characteristic are performed. The obtained results are compared to linear-prediction.
Convention Paper 6769 (Purchase now)
P18-7 An Ontology-Based Approach to Information Management for Music Analysis Systems—Samer Abdallah, Yves Raimond, Mark Sandler, Queen Mary, University of London - London, UK
We describe an information management system that addresses the needs of music analysis projects, providing a logic-based knowledge representation scheme for the many types of object in the domains of music and signal processing, including musical works and scores, performance events, human agents, signals, analysis functions, and analysis results. The system is implemented using logic-programming and semantic web technologies and provides a shareable resource for use in a laboratory environment. The whole is driven from a Prolog command line, where the use of Matlab as a computational engine enables experiments to be designed and run with the results being automatically stored and indexed into the information structure. We present as a case-study an experiment in automatic music segmentation.
Convention Paper 6770 (Purchase now)
P18-8 Pyramid Algorithm for the Restoration of Audio Signal Corrupted by Wideband Noise—Azaria Cohen, Itai Neoran, Waves - Tel Aviv, Israel
Restoration of noisy audio recordings necessitates maximum suppression of noise with minimum degradation of program material. Spectral suppression methods perform best with high frequency resolution but deliver poor performance with transients. While wavelet-based algorithms attempt to mitigate the time-frequency trade-off, they suffer from frequency aliasing. The suggested pyramidal algorithm is a good candidate for resolving the time-frequency resolution trade-off while avoiding aliasing. In this paper an algorithm for removal of wideband noise from old audio recordings is evaluated. The algorithm is based on the pyramidal algorithm and on a spectral method for noise suppression. Results show enhanced conservation of onsets with efficient reduction of noise. The algorithm is implemented in real-time.
Convention Paper 6771 (Purchase now)
P18-9 Digital Music Notation Transformation Using XML—Erich Christian Teppan, Harald Kosch, University Klagenfurt - Carinthia, Austria
The basic problem this paper deals with is how to convert western music notations, written for chromatic instruments, into special tablatures for diatonic instruments. There are just a few software programs addressing this problem but with the lack of full automatic operation and flexibility. This was the main reason for the development of new data formats and a new transformation algorithm, which are more suitable for the above-mentioned problem. Combined in an accurate software architecture, the newly-developed algorithm performs the transformation from a chromatic piece of music into a data format, which represents a diatonic tablature.
Convention Paper 6772 (Purchase now)
P18-10 A Service-Oriented High-Performance Architecture for Large Scale Audio Archives—Stephan Schneider, Blue Order AG - Kaiserslautern, Germany
This paper describes a solution for large audio archives that have been developed using a service-oriented architecture (SOA). The audio archiving system is designed as a framework of Web services that are controlled centrally by a workflow engine. The audio archiving system offers hierarchical storage, import, export, and conversion of audio files in various formats. Search and retrieval are based on textual metadata that are entered into an entity-relationship model (ERM). The archive offers a Web-based interface that can be used with a standard Web browser. Current installations are coping with several hundreds of users, more than 700,000 metadata entries and 16 TB of audio files.
Convention Paper 6773 (Purchase now)
P18-11 A Robust Music Retrieval System—Yuan-Yuan Shi, Xuan Zhu, Hyoung-Gook Kim, Ki-Wan Eom, Samsung Advanced Institute of Technology - HaiDian District, Beijing, China
A robust music audio fingerprinting system for automatic music retrieval is proposed in this paper. The fingerprinting feature is extracted from the long-term dynamic modulation spectrum estimation in the perceptual compressed domain. The modulation frequency analysis, smoothing with a low-pass filter and the low resolution quantization significantly improve the robustness of the feature. Further, the fast searching problem is solved by looking up hash table with 32-bit hash values. The hash value bits are quantized from the logarithmic scale modulation frequency coefficients. The system obtains 42.62 percent, 92.52 percent, 97.00 percent, or 99.67 percent search precision with approximately 3.0 percent false positive rate when the query clips’ signal-to-noise ratio is <0 dB, 0~5 dB, 5~15 dB, or >15 dB, respectively.
Convention Paper 6774 (Purchase now)
|(C) 2006, Audio Engineering Society, Inc.