v7.0, 20040922, me

Friday, October 29, 1:30 pm – 6:00 pm
Session I PSYCHOACOUSTICS, PERCEPTION, AND LISTENING TESTS

Chair: Gilbert Soulodre, Communications Research Centre, Ottowa, Ontario, Canada

1:30 pm
I-1 Preferred Listening Levels and Acceptance Windows for Dialog Reproduction in the Domestic Environment—Eric Benjamin, Dolby Laboratories, San Francisco, CA, USA
Audio reproduction systems have as their first goal the ability to reproduce the program sounds at the level desired by the user. There are a number of areas in audio system design where this is critical. The system should be able to reproduce audio at the levels needed by typical users, and presumably should also have the ability to accommodate a substantial portion of the desired reproduction levels. Beyond this simple requirement, some systems should have their behavior optimized for the level at which they will be used. The literature on listening levels has been surveyed, and new data has been gathered to determine what the preferred listening levels are for a variety of listening circumstances. Additional experiments have been done to estimate the range of listening levels which may be acceptable to the typical listener.
Convention Paper 6233

2:00 pm
I-2 Evaluation of Different Loudness Models with Music and Speech Material—Esben Skovenborg, University of Aarhus, Århus, Denmark, TC Electronic A/S, Risskov, Denmark; Søren H. Nielsen, TC Electronic A/S, Risskov, Denmark
The evaluation of twelve models of loudness perception is presented. One of the loudness models is based on a novel algorithm, and another is based on a combination of two known measurement techniques. The remaining models are all implementations of common or standardized loudness algorithms. The ability of each model to predict or measure the subjective loudness of speech and music segments is evaluated. The reference loudness is derived from two listening experiments using the speech and music segments as stimuli. Different statistical measures are employed in the evaluation of the models, so that both the absolute performance of the models and the performance relative to the between-listener disagreement are measured.
Convention Paper 6234

2:30 pm
I-3 Level Detection Tunings and Techniques for the Dynamic Range Compression of Audio Signals—Ryan Cassidy, Stanford University, Stanford, CA, USA
In a recent work, we reviewed the basics of level detection for dynamic range compression, and presented various tunings of level detector parameters for optimal correspondence with well-known and recently proposed facts pertaining to loudness perception. In this paper we review key points and present several extensions. We review the mathematics behind the operation of the popular root-mean-square detector, with special attention paid to the effect of time constants on detector performance. We compare an equal loudness filter, designed in the prior work to compensate for frequency-dependent steady–state loudness perception, to standard weightings. Updated results based on newly standardized loudness contours are also presented.
Convention Paper 6235

3:00 pm
I-4 A New Objective Measure of Perceived Loudness—Alan Seefeldt, Brett Crockett, Michael Smithers, Dolby Laboratories, San Francisco, CA, USA
The applications of an accurate objective measure of subjectively perceived loudness are many. Accordingly, the ITU-R has initiated a study to identify such a measure for a new recommendation. A new objective loudness measurement based on modifications to a traditional psychoacoustic model of perceived loudness was developed for this study. When compared to subjective loudness matching data generated outside the ITU-R, the new measure is found to perform better than simpler weighted power measurements and the unmodified psychoacoustic model.
Convention Paper 6236

3:30 pm
I-5 Training of Listeners for Evaluation of Spatial Attributes of Sound—Juha Merimaa, Wolfgang Hess, Ruhr-Universität Bochum, Bochum, Germany
A group of listeners were engaged in training to learn to evaluate auditory source width (ASW) and listener envelopment (LEV). The training consisted of discussions on the perception of spatial sound and visualization of both attributes with drawings. After each session the subjects evaluated a set of stimuli consisting of different source signals simulated in a few chosen acoustical environments. Most subjects developed consistent criteria for their judgments and maintained them throughout the training and a subsequent control experiment conducted two months later. However, considerable individual differences were found. Analysis of the data revealed that the large part of the differences was due to different judgments between the chosen source signals. The training also suggested that some differences could have been caused by the translation from the “spatial image” formed of the stimuli to the unidimensional judgments. A further experiment showed that this was not the case.
Convention Paper 6237

4:00 pm
I-6 Development of a Sound Quality Evaluation System—Preben Kvist, Carsten Thomsen, DELTA Acoustics & Vibration, Hørsholm, Denmark; Sangil Park, Joonhyun Lee, Samsung Electronics, Suwon, Korea; Finn Agerkvist, Technical University of Denmark, Lyngby, Denmark
This paper describes the development of the first version of the Sound Quality Evaluation System. The purpose of the system is to predict the subjective sound quality of home theater systems based on objective measurements. Sixteen home theater systems were measured in an anechoic room. Several metrics expected to correlate with the subjective quality were proposed and tested. A model for the sound quality was created by mapping the subjective evaluations of the home theater systems with the metrics calculated for each system. Correlation between subjective listening test and the prediction is presented.
Convention Paper 6238

4:30 pm
I-7 The Effect of Room Acoustics on MP-3 Audio Quality Evaluation—Daniël Schobben, Steven van de Par, Philips Research Laboratories, Eindhoven, The Netherlands
The impact of using loudspeaker versus headphone playback on the subjective quality of compressed audio is investigated. It is shown that reverberation—and to a lesser extent cross-talk, which both are introduced naturally in loudspeaker playback—can effectively hide coding artifacts. In double blind listening tests subjects had to rate MP3 coded excerpts at various bit rates. The excerpts were played back over headphones. Reverberation and cross-talk were introduced artificially to simulate loudspeaker playback so that their impact could be assessed separately. Results show that quality scores of the reverberated excerpts were significantly higher than for the corresponding “dry” excerpts for 64 kb/s bit rate while these differences diminished with increasing bit rate. This indicates that coding artifacts can become less audible in reverberant listening conditions.
Convention Paper 6239

5:00 pm
I-8 The Restitution of Timbre by Loudspeakers in a Listening Room: Perceptual and Physical Measurements—Mathieu Lavandier, Philippe Herzog, Sabine Meunier, Laboratoire de Mécanique et d’Acoustique, Marseille, France
This paper deals with the relationships between two parallel evaluations of a panel of loudspeakers: perceptual measurements and physical ones. The sound radiated by the loudspeakers has been recorded. The recordings were submitted to both listening tests and signal analysis. Pair-comparison tests were run using headphones, so the spatial dimension of sound reproduction is not investigated. This first attempt revealed two main perceptual dimensions. They were independent of the tested recording techniques and musical excerpts. We determined a suitable method of analysis for the physical measurements, and then we looked for objective attributes correlated with the perceptual dimensions.
Convention Paper 6240

5:30 pm
I-9 Subjective Test of Class D Amplifiers Without Output Filter—Finn Agerkvist, Technical University of Denmark, Lyngby, Denmark; Lars Fenger, Bang & Olufsen ICEpower a/s, Lyngby, Denmark
This paper presents subjective tests of the active transducers concept. The tests are designed to determine whether the output filter on class D amplifiers used in an active loudspeaker can be omitted without audible error occurring. The input signal of the amplifiers was limited to 0 to 3 kHz corresponding to that of a woofer unit. A listening panel of seven persons was used in the tests. The tests showed that no errors could be detected.
Convention Paper 6241