AURORA³ (Anechoic & Universal Research Observation Rooms for Audio, Acoustics & AI)

The UK acoustics industry directly accounts for over 16,000 jobs and £4.6 billion of GDP and contributes to vibrant growth sectors of the UK economy: communication technologies, consumer electronics, healthcare, and the creative industries. We are witnessing a transformation of this industry through two key technologies — spatial audio and machine learning (ML). Over the past decade, spatial audio has become a cornerstone of media platforms and user experiences, featured in technologies ranging from Dolby Atmos sound on Netflix and Apple AirPods, to second-generation virtual/augmented reality headsets, live music performances, and art installations. More recently, ChatGPT’s media explosion has brought the capabilities of artificial intelligence (AI) into the popular consciousness, heralding the ferocious potential of ML technologies. Despite exciting advancements in audio, acoustics, and AI, a main obstacle is stifling research—scarcity of data resources. This data can take several forms, from audio recordings under various acoustic conditions to human responses to auditory stimuli. Current infrastructure in the UK is not set up to efficiently capture such data, forcing the research community into relying either on large, simulated datasets which may not match real-world conditions, or on real-world datasets that have limited variability in terms of room acoustic conditions. This is problematic because the effectiveness of ML and spatial audio methods hinges on the availability of large (in the case of ML, vast) quantities of high-quality, diverse data collected under realistic acoustic conditions. The alternative, i.e. relocating people and audio equipment to record data in many different locations, is expensive and impractical. Embedded within the established research strengths around spatial sound perception and ML in audio at Surrey, AURORA³ (pronounced “aurora cubed” or simply “aurora”) aims to overcome this data resourcing challenge and establish next-generation strategic infrastructure. AURORA³ will combine state-of-the-art audio equipment and facilities for rigorous, yet precise control over a wide range of acoustic conditions: a variable acoustics room equipped with adjustable wall panels that can transform the acoustics of the room at the push of a button—from church-like reverberance to studio-like dryness—and a moving wall for modifying the room volume; an acoustic anechoic chamber tailored to spatial audio applications, featuring a semi-permanent spherical loudspeaker array for investigating interactions with simulated or recorded environments and soundscapes. Together, these facilities will enable full acoustic environment control for fast, accurate and reproducible physical and perceptual data collection. Technologies developed using AURORA³ will be more robust to different real-world environments and will be able to better model the underlying physical and perceptual phenomena involved in audio and acoustics. This will unlock benefits across a range of applications, including (a) more immersive Virtual/Augmented/Mixed Reality experiences for entertainment, metaverse, remote communications, and virtual prototyping; (b) more accurate and robust consumer electronics devices for automatic speech recognition, speech enhancement and source separation; and (c) smarter hearing aid devices that better understand and adapt to the acoustic scene.