Keynote: Speech and Audio Capabilities in Our Everyday Lives – Can it be Trusted?
Jennifer A. Williams · 2023
Long before we had the iPhone and Alexa, intelligent speech and audio R&D was rooted in defence applications. Historically, this stemmed from the need for secure communication channels after WWII and practical applications such as noise cancellation, speech translation, language identification, and the ability to verify a person's identity when speaking over the telephone. Many of the commercially available audio technologies that we use today stand on the shoulders of decades of research in signals processing. Audio technologies are commonplace and serve a variety of purposes in our everyday life. The big boom of neural networks last decade re-shaped audio processing, and new trends aim to create audio technologies that operate with ultra-low resources on edge devices. Microphones and speakers are everywhere. We are now at a tipping point where audio AI bleeds into many other research disciplines - from medicine and healthcare to autonomous vehicles and fake media. This talk will present some examples of interdisciplinary research and highlight recent and ongoing work at the University of Southampton, as well as current challenges and opportunities. Topics to be covered include: • Trust in audio AI systems • Safety, security, and privacy • Smart buildings and cities.