Auditory Timbre and Spatialisation: Signal Analysis and Perception of Source Widening

Abstract

In this work, auditory perception of source widening is examined in the context of different source signal timbre. Perception of widening of source or auditory source width (ASW) arises in three cases: (i) In the presence of reverberation, which has been referred to as reverberant source width (RSW); (ii) Distributed sources such as an ensemble, where multiple sources are physically placed widely, referred as ensemble source width (ESW); and (iii) In hearing disabilities, where localisation is poor in the presence of interfering sources and hence a widened or diffused source width (DSW) is perceived. Though the physics of the problem is different in each of the above cases, we observe that the perception of source widening occurs in all the three cases. We also show analytically that in the case of localised, reverb and ensemble sources located about a particular direction, binaural cross-correlation has interesting properties: (i) for localised source, cross-correlation is energy compact, followed by (ii) reverb source and (iii) ESW has highly dispersed cross-correlation compared to localised and reverb sources with the same angle of arrival and degree of decorrelation as reverb sources. Traditionally, (1-IACC) has been used as a measure for RSW and in the literature this measure is used for ESW also. We propose a combination of timbre-independent phase-based angular measure for the physical extent of the sources, localising all or many individual sources using HRIR correlation functions and timbre dependent mean time-bandwidth energy (MTBE) measure for relative perceptual weighting to compare ensemble of different timbres. This analysis gives rise to possible applications in ensemble rendering and insights into improving hearing aids for hearing impaired listeners. Frequency sensitivity to change in IACC, and hence ASW, has been studied using binaural presentation of modulated sinusoids. In this work, we observe a similar sensitivity to ESW by presenting listeners with spatially wide sources using narrow band noise signals. We observe that frequency sensitivity of ASW and ESW are similar. We also study bandwidth sensitivity and observe that with increase in bandwidth, the perceived width of the ensemble increases. We simulate ensemble-like music signals of different spectro-temporal distribution to probe the timbre dependency of human perception. The listeners are asked to rate the ESW of the simulated distributed sources. Broadly, music signals can be classified as sustained instruments, partially sustained, partially transient and predominantly transient signals. Low frequency sustained instruments give rise to a wider percept than semi-sustained and transient signals. We also explore the difference between discrete and continuous spectral sources for spatialisation. We observe that continuous spectra do give rise to stable mono- tonic width perception with change in physical width. On the other hand, in the case of discrete spectra, we do not perceive a stable monotonic perception. We developed a MUSHRA like (Multiple stimulus hidden reference and anchor) listening methodology for estimating the accuracy of direction perception of the target signals with and without interference by normal listeners. We observe that the accuracy of direction perception of the target is high without interference. In the presence of interference, we observe that the perceived target direction is away from that of interference, thus increasing the perceived angular separation. This perceptual effect may be used in the design of binaural hearing-aids to enhance binaural perception of localised sources in the presence of interference. Overall, in this work, we study the perceptual interaction of signal timbre and spatialisation in the perception of ensemble source width. We study the sensitivity of several parameters like frequency, bandwidth, spectro-temporal energy distribution and role of fine AM-FM parameters.