Periodic Reporting for period 3 - SONICOM (SONICOM - Transforming auditory-based social interaction and communication in AR/VR)
Periodo di rendicontazione: 2023-07-01 al 2024-12-31
Immersive audio is our everyday experience of being able to hear and interact with sounds around us. Simulating spatially located sounds in virtual or augmented reality (VR/AR) must be done in a unique way for each individual. This has been the object of extensive research in the past years but, several major challenges still exist this area, which require an improved understanding and modelling of the human spatial hearing mechanisms. Further, the impact of immersive audio beyond perceptual metrics such as presence and localisation is still an unexplored area of research, specifically when related to social interaction within virtual environments, entering the behavioural and cognitive realms.
SONICOM puts together some of the most important research centres across Europe to tackle these research challenges by using a multidisciplinary approach. Research involves a mixture of digital signal processing and artificial intelligence, with psychology and perceptual modelling; exploring new forms, levels, and dimensions of interaction within auditory-based immersive environments. Furthermore, in order to reinforce the idea of reproducible research and promoting future development and innovation in the area of auditory-based social interaction, the SONICOM Ecosystem will be created and shared with the wider community, which will include auditory data closely linked with model implementations and immersive audio rendering components.
SONICOM puts together some of the most important research centres across Europe to tackle these research challenges by using a multidisciplinary approach. Research involves a mixture of digital signal processing and artificial intelligence, with psychology and perceptual modelling; exploring new forms, levels, and dimensions of interaction within auditory-based immersive environments. Furthermore, in order to reinforce the idea of reproducible research and promoting future development and innovation in the area of auditory-based social interaction, the SONICOM Ecosystem will be created and shared with the wider community, which will include auditory data closely linked with model implementations and immersive audio rendering components.
SONICOM’s 3rd reporting period began with the participation in the Royal Society Summer Science Exhibition (London July 2023), and for which we have a physical and digital presence (https://www.axdesign.co.uk/royal-society-summer-science-exhibition(si apre in una nuova finestra)). SONICOM’s custom-designed stand exhibiited for a whole week and visited by more than 10,000 people.
In the last 10 months, WP1 and WP2 were at the core of SONICOM’s research. Several studies have been carried out and published within the 3 tasks of WP1; achieving significant results both in terms of general understanding of spatial hearing mechanisms (specifically focussed on the matters of personalisation and reverb perception), and in terms of developed methods and tools for virtual spatial audio rendering, from the measurement and synthesis of individual Head Related Transfer Functions (HRTFs), to their estimation using parametric pinna models and/or photogrammetry; from the rendering of directional sound sources features, to the matching of real reverberation within the virtual domain; and so on.
Similarly, several studies have been carried out in WP2, this time looking at higher-level perceptual processes, closer to the cognitive side of things. Beyond the flagship study on proxemics, which produced very interesting results on the relationship between the distance of a speaker and his/her perceived personality traits, other experiments have been designed and carried out for example, looking at interactions in complex simulated environments, as well as in AR scenarios; or exploring in depth the process of adaptation to non-individual HRTFs and how much the acquired training can be generalised to other filters, tasks and cues; or again looking at (and modelling) dynamic localisation, both in 3 and 6 Degrees of Freedom.
WP3 was also an essential part of RP3, having the task to take what was discovered and developed in WP1-2, implement it into a usable framework (the Binaural Rendering Toolkit - BRT), and deliver it to WP4 for the evaluations, and to WP5 for ensuring that all the tools and models will keep existing and, potentially, being developed after the end of SONICOM. Several implementations and versions of the BRT were released within the SONICOM consortium; several meetings and workshops took place to clarify requirements coming from the planned WP4 evaluations, and to ultimately train researchers in using such custom tools. The build of the self-personalising headphones also moved forward, and we now have a functional prototype, which has been successfully deployed within experimental settings.
WP4 effectively started within this reporting period; working towards the definition of the scenarios and significantly involving the SME partners in the design of the evaluations. We are now ready to start working on the implementation (this has already happened for some of the tasks), aiming at starting to carry out the actual experiments and installations in 2025.
WP5 was also very active. In addition to the release of several data, models, and tools through various existing channels, the SONICOM ecosystem design has now been completed and we are working on its implementation and, ultimately, its release. In addition to this, within RP3 we have launched and completed the first Listener Acoustic Personalisation (LAP) challenge (https://www.sonicom.eu/lap-challenge/(si apre in una nuova finestra)). This was a big success for SONICOM, as several EU and overseas participants were involved, and both the launch and the closing events were hosted within high-impact conference venues.
In the last 10 months, WP1 and WP2 were at the core of SONICOM’s research. Several studies have been carried out and published within the 3 tasks of WP1; achieving significant results both in terms of general understanding of spatial hearing mechanisms (specifically focussed on the matters of personalisation and reverb perception), and in terms of developed methods and tools for virtual spatial audio rendering, from the measurement and synthesis of individual Head Related Transfer Functions (HRTFs), to their estimation using parametric pinna models and/or photogrammetry; from the rendering of directional sound sources features, to the matching of real reverberation within the virtual domain; and so on.
Similarly, several studies have been carried out in WP2, this time looking at higher-level perceptual processes, closer to the cognitive side of things. Beyond the flagship study on proxemics, which produced very interesting results on the relationship between the distance of a speaker and his/her perceived personality traits, other experiments have been designed and carried out for example, looking at interactions in complex simulated environments, as well as in AR scenarios; or exploring in depth the process of adaptation to non-individual HRTFs and how much the acquired training can be generalised to other filters, tasks and cues; or again looking at (and modelling) dynamic localisation, both in 3 and 6 Degrees of Freedom.
WP3 was also an essential part of RP3, having the task to take what was discovered and developed in WP1-2, implement it into a usable framework (the Binaural Rendering Toolkit - BRT), and deliver it to WP4 for the evaluations, and to WP5 for ensuring that all the tools and models will keep existing and, potentially, being developed after the end of SONICOM. Several implementations and versions of the BRT were released within the SONICOM consortium; several meetings and workshops took place to clarify requirements coming from the planned WP4 evaluations, and to ultimately train researchers in using such custom tools. The build of the self-personalising headphones also moved forward, and we now have a functional prototype, which has been successfully deployed within experimental settings.
WP4 effectively started within this reporting period; working towards the definition of the scenarios and significantly involving the SME partners in the design of the evaluations. We are now ready to start working on the implementation (this has already happened for some of the tasks), aiming at starting to carry out the actual experiments and installations in 2025.
WP5 was also very active. In addition to the release of several data, models, and tools through various existing channels, the SONICOM ecosystem design has now been completed and we are working on its implementation and, ultimately, its release. In addition to this, within RP3 we have launched and completed the first Listener Acoustic Personalisation (LAP) challenge (https://www.sonicom.eu/lap-challenge/(si apre in una nuova finestra)). This was a big success for SONICOM, as several EU and overseas participants were involved, and both the launch and the closing events were hosted within high-impact conference venues.
In the last 4 years, SONICOM has substantially impacted the academic communities by heavily contributing to the design and evaluation of novel immersive audio technologies and techniques, as well as to the overall advancement of knowledge and understanding in the spatial acoustics and immersive audio research domains. Beyond out work on parametric pinna models, which has already impacted several novel research within and outside the consortium, our work on perceptual- and numerical-based HRTF selection frameworks has moved forward to the mapping of existing approaches across the research community, and is already contributing to consolidating their use within and beyond SONICOM. Explorative work has been completed and published looking at the interactions between HRTF choices and speech intelligibility; this is a relatively unexplored research area which has the potential to generate major impact on our understanding of immersive audio communication and interaction within AR/VR. Considering specifically the AR context, where real and virtual combine in a single domain, SONICOM’s work focussed on 3 studies looking at matching the reverberation of the real environment with the virtual, and at exploring which computational/rendering choices (including machine-learning-based ones) and parameters are relevant from a perceptual point of view. Results from these and other experiments have been (or will soon be) published and will allow us to achieve a seamless blend between the real and virtual auditory domains.
In addition, significant work has been completed mapping, planning, and executing activities aimed at exploring and modelling how the physical characteristics of spatialised auditory stimuli can influence observable behavioural, physiological, kinematic, and psychophysical reactions of listeners within AR/VR-based social interaction scenarios. The SONICOM 3D Speaker Personality Corpus will be soon released, and will hopefully become a standard tool within the academic community for exploring interactions between the simulated distance/position of speakers and the traits that listeners attribute to them. Furthermore, our study on proxemics has now shown significant results, which will be soon published showing how AI-based models can be trained to infer socially and psychologically relevant perceptions of a user from the widest possible array of measurable aspects in an AR/VR environment.
In addition, significant work has been completed mapping, planning, and executing activities aimed at exploring and modelling how the physical characteristics of spatialised auditory stimuli can influence observable behavioural, physiological, kinematic, and psychophysical reactions of listeners within AR/VR-based social interaction scenarios. The SONICOM 3D Speaker Personality Corpus will be soon released, and will hopefully become a standard tool within the academic community for exploring interactions between the simulated distance/position of speakers and the traits that listeners attribute to them. Furthermore, our study on proxemics has now shown significant results, which will be soon published showing how AI-based models can be trained to infer socially and psychologically relevant perceptions of a user from the widest possible array of measurable aspects in an AR/VR environment.