XReverb: efficient generation and perceptual evaluation of reverberant spatial audio for Extended Reality

Descrizione del progetto

Un nuovo modello di realtà estesa per la localizzazione del suono nel riverbero

Il termine realtà estesa si riferisce alla tecnologia che consente agli utenti di immergersi in un ambiente 3D popolato da oggetti virtuali percepibili attraverso i sensi (generalmente vista, udito e tatto). Solitamente, il termine viene usato in modo generico per indicare realtà virtuale, aumentata e mista. La realtà estesa è impiegata in numerosi settori, che spaziano dalla produzione, dalla formazione e dall’istruzione alla salute e all’intrattenimento. Il progetto XReverb, del programma di azioni Marie Skłodowska-Curie (MSCA), ha concentrato la propria attenzione sulla realtà estesa uditiva ed esaminerà il modo in cui le persone percepiscono il riverbero spaziale per costruire metodi di rendering più efficienti che consentano di ottenere un audio spaziale di elevata qualità nella realtà estesa. Nello specifico, il progetto ideerà una nuova tecnica di codifica del riverbero basata su un quadro ambisonico di ordine variabile e un modello uditivo per la localizzazione del suono nel riverbero.

Obiettivo

Extended Reality (XR) refers to the technology which creates a 3D immersive environment where a user can perceive and interact with virtual objects by means of a head mounted display. Even though XR is still an active research area, its European market has already reached an estimated worth of €34 million, employing up to 480,000 people. It is expected that a mature XR infrastructure will raise the standards for remote working by enabling functional virtual workspaces, which could stimulate the European economy by offering equal opportunities to workers regardless of geography, and reduce greenhouse gas emissions due to less commuting, which is in line with European goals. A key part of XR is spatial audio, or sound signals which the user perceives as if they come from specific locations in a 3D space. Realistic spatial audio can be computationally expensive, particularly when simulating reverberation, i.e. the interaction of a sound source and the environment. This project investigates how humans perceive spatial reverberation to build more efficient rendering methods which enable high-quality spatial audio on XR. To that end, a novel reverberation encoding technique is proposed, based on a variable order Ambisonics framework, and a model for sound localisation in reverberation will be developed, which is expected to be a significant advancement in the field of auditory perception. The project will take place at a leading multidisciplinary group specialised on real-time spatial audio processing and who can offer excellent international collaboration opportunities. The researcher will bring skills on evaluation of spatial reverberation and inter-sectoral experience from having worked at a world leading research group within the industry, which will facilitate the transfer of knowledge. The proposed work will expand the researcher’s experience, competencies and professional networks, potentiating the development of his career as an independent researcher.