9:00 am

H-1 Subjective Evaluation of an Audio Distribution Coding System

Dane Grant, Grant Davidson and Louis D. Fielder, Dolby Laboratories, Inc., San Francisco, CA, USA

Subjective quality is a critical indicator of the suitability of bit-rate reduction codecs for digital audio contribution/distribution applications. Accordingly, a formal double-blind test was conducted to evaluate the subjective quality of a contribution/distribution cascade (eight Dolby E audio codecs in tandem), both separately and when combined with three standardized stereo emission codecs. In this test, the contribution/distribution cascade exceeded the ITU-R basic audio quality requirements for broadcast applications. The results further indicate that when using the contribution/distribution cascade in tandem with any one of the emission codecs, audio quality is effectively limited by the emission codec alone. The test methodologies conformed to Recommendation ITU-R BS.1116, and hence represent an estimate of worst-case performance.

Convention Paper 5443

9:30 am

H-2 Auditory Models in Audio Coding

João Manuel Rodrigues, Ana Maria Tomé and Tomás Oliveira e Silva, Universidade de Aveiro/IEETA, Aveiro, Portugal

Two strategies have been used to evaluate the perceptual significance of a distortion introduced into a signal: the masking threshold concept, and the internal representation approach. Psychoacoustic models found in the perceptual audio coding literature generally follow the first approach in spite of the recognized unsuitability of this approach for the task. The more plausible internal representation approach is currently exploited in objective measurement of audio quality but not, to our knowledge, in audio coders. In this paper, we explore a standard internal representation model, and find some statistical relations that might, in the future, be applied to audio coding.

Convention Paper 5444

10:00 am

H-3 The Reduction of Distortion in the Dynamic Compressor

Piotr Kleczkowski, Technical University of Mining and Metallurgy, Crakow, Poland

The action of the dynamic compressor introduces non-linear distortion. This is meaningful for the attack portion of this action. The distortion is analyzed and the feasibility of its reduction is investigated. It is possible to reduce the distortion by the appropriate choice of the time function actually controlling the gain. It is shown that the distortion can be measured in absolute scale, but it is difficult to develop a psychoacoustically justified measure. Some listening tests have been performed and their results are compared to quantitative analysis leading to interesting conclusions.

Convention Paper 5445

10:30 am

H-4 Binaural Modeling of Multiple Sound Source Perception: Methodology and Coloration Experiments

Kazuho Ono, NHK Science and Technical Research Laboratories, Setagaya-ku, Tokyo, Japan
Ville Pulkki and Matti Karjalainen, Helsinki University of Technology, Espoo, Finland

Binaural modeling of coloration perceived due to multiple coherent sources is studied under the condition that sounds arrive at a listener successively within a certain time delay. The model simulates the perception of coloration of two horizontally located sources under prominent precedence effect conditions. Listening experiments are conducted to ensure the validity of the modeling. A new methodology is adopted in the experiments to minimize the error owing to individuality of HRTFs and inaccuracy of HRTF measurements.

Convention Paper 5446