Hagen Wierstorf

Spatial perception of sound fields recorded by spherical microphone arrays with varying spatial resolution

The area of sound field synthesis has significantly advanced in the past decade, facilitated by the development of high-quality sound-field capturing and re-synthesis systems. Spherical microphone arrays are among the most recently developed systems for sound field capturing, enabling processing and analysis of three-dimensional sound fields in the spherical harmonics domain. In spite of these developments, a clear relation between sound fields recorded by spherical microphone arrays and their perception with a re-synthesis system has not yet been established, although some relation to scalar measures of spatial perception was recently presented. This paper presents an experimental study of spatial sound perception with the use of a spherical microphone array for sound recording and headphone-based binaural sound synthesis. Sound field analysis and processing is performed in the spherical harmonics domain with the use of head-related transfer functions and simulated enclosed sound fields. The effect of several factors, such as spherical harmonics order, frequency bandwidth, and spatial sampling, are investigated by applying the repertory grid technique to the results of the experiment, forming a clearer relation between sound-field capture with a spherical microphone array and its perception using binaural synthesis regarding space, frequency, and additional artifacts. The experimental study clearly shows that a source will be perceived more spatially sharp and more externalized when represented by a binaural stimuli reconstructed with a higher spherical harmonics order. This effect is apparent from low spherical harmonics orders. Spatial aliasing, as a result of sound field capturing with a finite number of microphones, introduces unpleasant artifacts which increased with the degree of aliasing error.

Binaural Assessment of Multichannel Reproduction

This chapter outlines the problem of evaluating multichannel reproduction by example of the wave-field synthesis method. This method is known for providing good localization of reproduced source within an extended listening area. The localization performance for a virtual point sources was investigated for various listener positions and loudspeaker-array configurations. Respective results of listening-test were compared with localization predictions by a binaural model. With this model, a localization map can be obtained that covers most listener positions within the synthesis area. With such a localization map, designers of loudspeaker-setups for wave-field synthesis can estimate the localization and localization accuracy to be expected from a given multichannel setup. To enable perception of sound sources at arbitrary positions within the synthesis area of a given wave-field synthesis implementation, input signals to the two ears had to be generated. This was realized by means of dynamic binaural synthesis, a technique that allows for instantaneous switching between different listening scenarios. In a formal pre-test, it was verified that dynamic binaural simulation has no influence on the listeners’ localization performance as compared to natural hearing. Both the test procedure and the modeling results can be taken as a basis for further research regarding the evaluation of multichannel reproduction, an area that is still sparsely covered.

Evaluation of perceptual properties of phase-mode beamforming in the context of data-based binaural synthesis

Several approaches to data-based binaural synthesis have been published that process the sound field captured by a spherical microphone array. The captured sound field is typically decomposed into plane waves which are then auralized using head-related transfer functions. The decomposition into plane waves is often based upon modal beamforming techniques which represent the sound field with respect to surface spherical harmonics. The achievable spatial bandwidth is limited due to practical considerations. This paper investigates the perceptual implications of these limitations in the context of data-based binaural synthesis.

Assessing localization accuracy in sound field synthesisa)

Sound field synthesis methods like Wave Field Synthesis (WFS) and Near-Field Compensated Higher Order Ambisonics synthesize a sound field in an extended area surrounded by loudspeakers. Because of the limited number of applicable loudspeakers the synthesized sound field includes artifacts. This paper investigates the influence of these artifacts on the accuracy with which a listener can localize a synthesized source. This was performed with listening tests using dynamic binaural synthesis to simulate different sound field synthesis methods and incorporated several listening positions. The results show that WFS is able to provide good localization accuracy in the whole listening area even for a low number of loudspeakers. For Near-Field Compensated Higher Order Ambisonics the achievable localization accuracy of the listener depends highly on the Ambisonics order and shows large localization deviations for low orders, where splitting of the perceived sound source was sometimes reported.

Open science in the Two!Ears project—Experiences and best practices

Two!Ears was an EU funded project for binaural auditory modeling with ten international partners involved. One of the project goals was to follow an Open Science approach in all stages. This turned out to be a challenging task as the project involved huge amounts of software, acoustical measurements, and data from listening tests. On the other hand, it was obvious from the positive experience with the Auditory Modelling Toolbox that an Open Science approach would have a positive impact and foster progression afterwards. As there existed no ready solution to achieve this goal at the beginning of the project, different paths for data management were tested. It was especially challenging to provide a solution for data storage. Here, the goal was not only the long term accessibility of the data, but also the revision control of public and private data for the development inside the project. In the end, the project was able to make most of its software and data publicly available, but struggled to apply the re...

Update on sound quality assessment with TWO!EARS

The paper summarizes the different test and modeling campaigns carried out in the EC-funded FET-Open project TWO!EARS (www.twoears.eu) for sound quality and Quality of Experience (QoE) evaluation of spatial audio reproduction technology like stereophony or Wave-field Synthesis (WFS). This work represents one of the two proof-of-concept application domains of the interactive listening model developed in TWO!EARS. One stream of our sound-quality-related work focused on listening tests and model development for the individual sound quality features localization and coloration. After briefly reviewing the modeling approaches for these individual features presented in more depth elsewhere, the paper presents data and modeling considerations for a set of pairwise preference listening tests, following a dedicated audio mixing and reproduction paradigm. For subsequent model development, the results are analyzed in different ways, for example in terms of the pairwise preference data directly, using the Bradley-Ter...

Azimuthal localization in 2.5D near-field-compensated higher order ambisonics

Sound Field Synthesis approaches aim at the reconstruction of a desired sound field in a defined target region using a distribution of loudspeakers. Near-Field Compensated Higher Order Ambisonics (NFCHOA) is a prominent example of such techniques. In practical implementations different artifacts are introduced to the synthesized sound field: spatial aliasing is caused by the non-zero distance between the loudspeakers. Modal bandwidth limitation is a well-established approach to reduce spatial aliasing in 2.5D NFCHOA, but introduces temporal and spectral impairments to the reproduced sound field which strongly depend on the relative position to the center of modal expansion. Also, the dimensionality mismatch in a 2.5D synthesis scenario results in a different amplitude decay compared to the desired sound field. Listening experiments already investigated the azimuthal localization in 2.5D NFCHOA. It is however unclear, in how far individual artifacts caused by spatial sampling, modal bandwidth limitation, a...

Recent advancements in massive multi-channel auralization

Massive multi-channel auralization approaches like Wave Field Synthesis and Higher Order Ambisonics experienced a pronounced hype in the late 2000s during which the primary research goal was maximizing the physical accuracy of the synthetic sound fields that they create. The hype eventually faded as the achievable advancements turned out to be limited due to fundamental restrictions. Though, activities are still being pursued in the domain with the focus shifted towards perception of synthetic sound fields. This talk gives an overview over current activities, which aim at understanding localization, timbre, and spatial impression in general. The results show that localization performance in synthetic sound fields is close to the performance in real sound fields. Timbre and spatial impression exhibit impairments that are directly linked to the physical limitations of the employed systems. Promising options for improvements regarding the synthesis of artificial reverberation are discussed.

Spatial Audio Rendering

Complementary to non-spatialized signals and their transmission, this chapter gives an overview of the quality of rendering methods that create spatial sound. Common methods and the underlying concept of a virtual sound scene are introduced and the herewith associated quality features. In particular, evaluation strategies and experimental results are presented in order to discuss spatial and timbral quality features of spatial audio rendering.

Representation of Intelligible and Distorted Speech in Human Auditory Cortex

The aim of the study is to identify cortical regions that process the meaning of spoken language, irrespective of acoustical realization, by means of functional MRI. Distorted speech material is used along with clean speech to allow for a controlled manipulation of intelligibility, depending on the presentation order. Speech distortion was performed by rotating spectral features. Psychophysical experiments were performed with 230 sentences of a speech intelligibility test in German language, to explore the effect of presentation order of clean and distorted speech. Results from 15 listeners indicate that distorted signals can reliably be presented either in a way well to understand, or completely unintelligible, depending on the exact presentation relative to the other speech material. The results were used to design a functional MRI experiment that allowed us to image brain activation for identical acoustic stimuli, only differing in their intelligibility. The fMRI data indicate that speech stimuli in general cause bilateral activation in the temporal lobes. The contrast between natural and distorted speech only comes up in regions beyond primary auditory areas, mainly in superior temporal gyrus. The contrast between acoustically identical stimuli only differing in their intelligibility reveals several areas, largely in the left hemisphere and clearly separate from the temporal lobes. These regions are interpreted to represent the processing of the meaning of intelligible speech, irrespective of their exact acoustical properties.