v7.0, 20040922, me
Sunday, October 31, 1:00 pm 3:00 pm
Session Z8 Posters: SIGNAL PROCESSING, Part 2
NOTE: During the first 10 minutes of the session all authors will present a brief outline of their presentation.
Z8-1 Increased Correlation in Blind Audio Watermark DetectionA Blessing in Disguise?Krishna Kumar S., Indian Institute of Science, Bangalore, India, and Centre for Development of Advanced Computing, Trivandrum, India; Thippur Sreenivas, Indian Institute of Science, Bangalore, India
Audio watermarks are often made signal-dependant to keep them inaudible in the host signals. Blind watermark detectors, which do not have access to the unwatermarked signal, may seem handicapped, because an approximate watermark has to be re-derived from the watermarked signal. Referring to the exact watermark scenario as semi-blind detector, some reduction in performance is anticipated in blind detection over semi-blind detection. The present paper is an experimental investigation into this issue, applied to a typical correlation-based audio watermark detection scheme. It is found, surprisingly, that the statistical performance of the blind detector is better than that of the semi-blind detector. It is found that the rederived watermark is better correlated to the host signal and hence leads to better detection performance. It is confirmed that this happens only if the embedded watermark is the same as the examined watermark.
Convention Paper 6309
Z8-2 Comparison of Effectiveness of Musical Sound Separation Algorithms Employing Neural NetworksBozena Kostek, Marek Dziubinski, Piotr Dalka, Gdansk University of Technology, Gdansk, Poland
In this paper several algorithms are presented, developed for musical sound separation. The proposed techniques for the decomposition of mixed sounds are based on the assumption that pitch of the sounds contained in the mix is known, i.e., inputs of the algorithms are pitch tracks of the signals contained in the mixture. The estimation process of phase and amplitude contours representing harmonic components is based on the limited number of inner product operations, performed on the signal with the use of complex exponentials matching pitch characteristics of the separated signals, and not on the discrete spectral representations calculated via DFT. In this paper examples of separation results are presented and each algorithm performance is analyzed. The effectiveness of separation algorithms consists in calculation of feature vectors (FVs) derived from musical sounds after the separation process is performed and then in feeding them the Neural Network (NN) for automatic musical sound identification. The experimental results are shown and discussed. A comparison of effectiveness of all presented algorithms is also included, and conclusions are derived.
Convention Paper 6310
Z8-3 Distortion Audibility in Inverse FilteringScott Norcross, Gilbert Soulodre, Michel Lavoie, Communications Research Centre, Ottawa, Ontario, Canada
Previous studies have shown that certain inverse filtering methods introduce audible artifacts that can degrade the audio signal. To correct some of these artifacts various techniques such as regularization, smoothing, and increasing the length of the inverse filter have been proposed. While these methods help in some cases they may also produce other artifacts or distortions that degrade the audio quality. In the present study formal subjective tests were conducted to systematically investigate modeled distortions similar to those found in inverse filtering. Parameters of the distortions, such as spectral shape, length, and time profiles were varied for the subjective tests. The results of the tests can be used to better understand the audibility of these artifacts and to create a perceptual model that can be used to design subjectively improved inverse filters.
Convention Paper 6311
Z8-4 Harmonic Sound Source Separation Using FIR Comb FiltersMikel Gainza, Dublin Institute of Technology, Dublin, Ireland; Bob Lawlor, National University of Ireland, Maynooth, Ireland; Eugene Coyle, Dublin Institute of Technology, Dublin, Ireland
A technique for separating harmonic sound sources using FIR comb filters is presented. First, a preprocessing task is performed by a multipitch estimator to detect the pitches that the signal is composed of. Then, a method based on the Short Time Fourier Transform (STFT) is utilized to interactively extract the harmonics belonging to a given source by using FIR comb filters. The presented approach improves upon existing sinusoidal model approaches in terms of the perceptual quality of the extracted signal.
Convention Paper 6312
Z8-5 AES Technical Committee on Signal Processing Educational CD ProjectRob Maher, Montana State University, Bozeman, MT
The AES Technical Committee on Signal Processing is developing a compact disc with educational material and demonstrations intended for students, educators, and working digital audio engineers. The material includes examples of quantization and dither, basic psychoacoustics, and practical DSP. The multi-mode CD will have both audio tracks and a CD-ROM section with a web-browser interface. The CD will be produced for sale by the AES Publications office.
Convention Paper 6313