Showing papers by "AG Armin Kohlrausch published in 2006"

Perceptual evaluation of music similarity.

Abstract: This paper presents an empirical method for assessing music similarity on a set of stimuli using triadic comparisons in a balanced incomplete block design. We first evaluated the consistency of subjects in their rankings and then the concordance across subjects. The concordance was also evaluated for different subject populations to assess the influence of experience of the subject with the musical material. We finally analysed subjects’ ranking by the means of multidimensional scaling. Similarity judgments were found to be rather concordant across subjects. Significant differences between musicians and non-musicians and between subjects being familiar or non-familiar with the music were found for a small number of cases. Multidimensional scaling reveals a proximity of songs belonging to the same genre, congruent with the idea of genre being a perceptual dimension in subjects’ similarity ranking.

Structural boundary perception in popular music.

Abstract: The automatic extraction of musical structure from audio is an important aspect for many music information retrieval (MIR) systems. The criteria on which structural elements in music are defined in MIR systems is often not clearly stated but typically stem from (music) theoretical or signal-based properties. In many cases, however, perceptual-based criteria are the most relevant and systems need to be trained on or modeled after the perception of structural elements in music. Here, we investigate the perception of structural boundaries to Western popular music and examine the musical cues responsible for their perception. We make links to music theoretical descriptions of structural boundaries and to computational methods for extracting structure. The methods and data presented here are useful for developing and training systems for the automatic extraction of musical structure as it is perceived by listeners.

Human performance in detecting audio-visual asynchrony.

Abstract: The perception of audio‐visual delays by human observers is typically characterized by two quantities: the point of subjective equality (PSE) and the sensitivity to asynchrony. The PSE can be derived from a temporal order judgment (TOJ) procedure or from the outcomes of a synchronous‐successive response paradigm. Sensitivity to asynchrony is derived from the steepness of the response curve in the TOJ paradigm. In this contribution we present data that show that PSE estimates derived from TOJ measurements are much more variable across observers than those based on synchronous‐successive data. Two synchronous‐successive methods were used with different response categories: (1) two categories: asynchronous, synchronous, and (2) three categories: audio first, synchronous, video first. Both synchronous‐successive methods yielded similar results. A slight influence of stimulus type on PSE estimates derived from synchronous‐successive data was observed. To conclude, we analyzed discriminability values obtained w...

On the influence of interaural differences on temporal perception of masked noise bursts.

Abstract: In this research, the influence of interaural differences on temporal positioning of the perceived onset of a dichotic broadband noise target in a diotic broadband noise masker was explored. Interaural time or level differences, a combination of these differences, and filtering with head-related transfer functions were applied to broadband noise bursts, to establish lateralization in quiet at center and right positions in the lateral plane. While applying these interaural differences, the subjects’ ability to accurately align the noise’s onset to the meter of a regular series of diotic reference onsets was investigated for various signal-to-noise ratios. At high target sensation levels, mean onset positions were close to physical isochrony and no systematic differences in temporal positioning due to different interaural cues were observed. At low target sensation levels, positioning accuracy decreased with a decreasing signal-to-noise ratio, and dichotic targets were positioned less accurately than diotic targets. This reduced temporal positioning accuracy could not be attributed to differences in detectability of the targets just above the threshold, nor to a reduction of effective target bandwidth of dichotic broadband targets. From these results, the reduced accuracy of temporal perception in binaural unmasking conditions may be seen as a consequence of temporal integration in binaural processing.

Using multidimensional scaling to determine distances in the perceptual space of rolling sounds.

Abstract: We recorded sounds generated by a ball rolling over a wooden plate. Ball diameter, rolling speed, and plate thickness were varied. In a difference‐scaling experiment, the important dimensions of the space spanned by these sounds were determined. All sounds were combined in pairs and subjects were asked to quantify the difference they heard in each pair. In this way all possible perceptual differences could play a role in the difference scalings without a need to interpret these differences in terms of underlying object properties such as size or speed. Using multidimensional scaling, a good fit for the scaled differences was obtained with a minimum of three dimensions. None of the dimensions corresponded directly with one of the three mechanical parameters varied in the rolling sounds. Differences in speed lead to a smaller distance in the thus‐obtained perceptual space than differences in size, for the ranges we used. This corresponds to results from experiments where the subjects were directly asked to ...

Using psychometric functions to determine auditory perception of the size and speed of a rolling ball.

Abstract: The purpose of the presented experiments is to quantify exactly how precise listeners can judge certain mechanical parameters of a system by listening to the sounds generated by this system. We used sounds generated by a ball rolling over a plate. In the recording of these sounds, the ball diameter, rolling speed, and plate thickness were varied. Subjects were asked to judge each of these three parameters in the sound in two different ways; in a paired comparison, two alternative forced choice, and using absolute magnitude estimation. Both methods were used to calculate psychometric functions which estimate the probability that a listener correctly identifies the difference in one of the physical parameters. Both paired comparison and absolute magnitude estimation lead essentially to the same psychometric functions. For each pair of sounds the percentage of correct responses is plotted against the fraction of their sizes. These datapoints are then fitted to a cumulative normal distribution. The good fit o...