Search

To auralize a room's acoustics in six degrees-of-freedom virtual reality (VR), a dense set of spatial room impulse response (SRIR) measurements is required, so interpolating between a sparse set is desirable. This paper studies the auralization of room transitions by proposing a baseline interpolation method for higher-order Ambisonic SRIRs and evaluating it in VR. The presented method is simple yet applicable to coupled rooms and room transitions. It is based on linear interpolation with RMS compensation, although direct sound, early reflections, and late reverberation are processed separately, whereby the input direct sounds are first steered to the relative direction-of-arrival before summation and interpolated early reflections are directionally equalized. The proposed method is first evaluated numerically, which demonstrates its improvements over a basic linear interpolation. A listening test is then conducted in six degrees-of-freedom VR, to assess the density of SRIR measurements needed in order to plausibly auralize a room transition using the presented interpolation method. The results suggest that, given the tested scenario, a 50-cm to 1-m inter-measurement distance can be perceptually sufficient.

Binaural rendering of Ambisonic signals is one of the most accessible ways of experiencing spatial audio. However, due to technical constraints, the rendering algorithm needs special care and advanced signal processing, especially for low Ambisonic orders. Next to more intricate parametric model-based approaches, other computationally efficient algorithms have emerged that provide powerful options. One particularly effective technique is the idea to discard the phase of HRTFs above a certain frequency limit, where the auditory system is less sensitive to phase information and instead utilize the available low order resolution to achieve an optimized magnitude response. This technique is known as the magnitude-least-squares (magLS) binaural rendering algorithm and often implemented as a recursive solution over frequency bins. However, altering the phase can lead to group delay errors, and therefore, frequency dependent misalignment, i.e. dispersion, of the HRIR. This issue is particularly prevalent with measurements that show a significant pre-delay, such as linear phase HRTFs. Besides analyzing the phase behavior of magLS, we present an effective way to preserve the group delay by continuing the phase over frequency as observed in the lower frequency region unaffected by the phase modification. This simple modification leads to further improvements for the magLS binaural Ambisonics rendering. Non

Ambisonics is a scene-based spatial audio format that has several useful features compared to object-based formats, such as efficient whole scene rotation and versatility. However, it does not provide direct access to the individual source signals, so that these have to be separated from the mixture when required. Typically, this is done with linear spherical harmonics (SH) beamforming. In this paper, we explore deep-learning-based source separation on static Ambisonics mixtures. In contrast to most source separation approaches, which separate a fixed number of sources of specific sound types, we focus on separating arbitrary sound from specific directions. Specifically, we propose three operating modes that combine a source separation neural network with SH beamforming: refinement, implicit, and mixed mode. We show that a neural network can implicitly associate conditioning directions with the spatial information contained in the Ambisonics scene to extract specific sources. We evaluate the performance of the three proposed approaches and compare them to SH beamforming on musical mixtures generated with the musdb18 dataset, as well as with mixtures generated with the FUSS dataset for universal source separation, under both anechoic and room conditions. Results show that the proposed approaches offer improved separation performance and spatial selectivity compared to conventional SH beamforming., Comment: Code and listening examples: https://github.com/francesclluis/direction-ambisonics-source-separation

Velvet noise is a sparse ternary pseudo-random signal containing only a small portion of non-zero values. In this work, the derivation of the spectral properties of velvet noise is presented. In particular, it is shown that the original velvet noise is white, i.e. has a constant power spectrum. For velvet noise variants with altered probability of polarity, the spectral characteristics are analytically derived. Crushed additive velvet noise is shown to have potential in the design of coloured sparse noise sequences, which are useful in acoustic signal processing.

openaire: EC/H2020/812719/EU//VRACE A common aim in virtual reality room acoustics simulation is accurate listener position dependent rendering. However, it is unclear whether a mismatch between the acoustics and visual representation of a room influences the experience or is even noticeable. Here, we ask if listeners without any special experience in echolocation are able to identify their position in a room based on the acoustics alone. In a first test, direct comparison between acoustic recordings from the different positions in the room revealed clearly audible differences, which subjects described with various acoustic attributes. The design of the subsequent experiment allows participants to move around and explore the sound within different zones in this room while switching between visual renderings of the zones in a head-mounted display. The results show that identification was only possible in some special cases. In about 74% of all trials, listeners were not able to determine where they were in the room. The results imply that audible position dependent room acoustic rendering in virtual reality may not be noticeable under certain conditions, which highlights the importance of evaluation paradigm choice when assessing virtual acoustics.

openaire: EC/H2020/812719/EU//VRACE Acoustically transparent head-worn devices are a key component of auditory augmented reality systems, in which both real and virtual sound sources are presented to a listener simultaneously. Head-worn devices can exhibit a high transparency simply through their physical design but in practice will always obstruct the sound field to some extent. In this study, a method for predicting the perceptual transparency of head-worn devices is presented using numerical analysis of device measurements, testing both coloration and localization in the horizontal and median plane. Firstly, listening experiments are conducted to assess perceived coloration and localization impairments. Secondly, head-related transfer functions of a dummy head wearing the head-worn devices are measured, and auditory models are used to numerically quantify the introduced perceptual effects. The results show that the tested auditory models are capable of predicting perceptual transparency and are therefore robust in applications that they werenot initially designed for.

New conference formats are emerging in response to COVID-19 and climate change. Virtual conferences are sustainable and inclusive regardless of participant mobility (financial means, caring commitments, disability), but lack face-to-face contact. Hybrid conferences (physical meetings with additional virtual presentations) tend to discriminate against non-fliers and encourage unsustainable flying. Multi-hub conferences mix real and virtual interactions during talks and social breaks and are distributed across nominally equal hubs. We propose a global multi-hub solution in which all hubs interact daily in real time with all other hubs in parallel sessions by internet videoconferencing. Conference sessions are confined to three equally-spaced 4-h UTC timeslots. Local programs comprise morning and afternoon/evening sessions (recordings from night sessions can be watched later). Three reference hubs are located exactly 8 h apart; additional hubs are within 2 h and their programs are aligned with the closest reference hub. The conference experience at each hub depends on the number of local participants and the time difference to the nearest reference. Participants are motivated to travel to the nearest hub. Mobility-based discrimination is minimized. Lower costs facilitate diversity, equity, and inclusion. Academic quality, creativity, enjoyment, and low-carbon sustainability are simultaneously promoted. Version of record

openaire: EC/H2020/812719/EU//VRACE Multichannel auralizations based on spatial room impulse responses often employ sample-wise assignment of an omnidirectional response to form loudspeaker responses. This leads to sparse impulse responses in each reproduction loudspeaker and the auralization of transient signals can sound rough. Based on this observation, we conducted a listening test to examine the general phenomenon of roughness due to spatial assignment. First, participants assessed the roughness of both Gaussian noise and velvet noise, assigned sample-wise to up to 36 loudspeakers by two algorithms. The first algorithm assigns channels merely by selecting random indices, while the second one constrains the time between two peaks on each channel. The results show that roughness already occurs when few channels are used and that the assignment algorithm influences it. In a second experiment, virtualizations of the test were used to examine the factors contributing to increased roughness. We systematically show the effect of spatial assignment on noise and conclude that besides time-differences, level-differences caused by head-shadowing are the principal cause for the perceived roughness. The results have significance in spatial room impulse response rendering and spatial reverberator design.