Recent advances in AI, embedded systems, and physical sensing are reshaping how machines perceive and interact with the world. Among sensing modalities, acoustics offers a distinctive combination of spatial richness, material penetrability, and passive observability. In this talk, Prof. Nirupam Roy will present a vision for Acoustic Intelligence—a new class of sensing systems that harness the physics of sound, structural design, and neural modeling to achieve fine-grained environmental understanding. He will highlight recent systems from his lab: ultra-low-power spatial sensors (e.g., Owlet, SPiDR) that use 3-D-printed acoustic microstructures to estimate direction and depth with high accuracy and minimal energy; neural architectures (e.g., SING) that integrate directional audio cues into language models for spatially grounded speech understanding; and privacy-preserving audio front-ends (e.g., VoiceSecure) that give users fine-grained control over how their voice data is interpreted and shared in real-time communication. These projects exemplify a broader shift toward embedding intelligence into the physical substrate of sensing—enabling capabilities previously unattainable with conventional approaches. Prof. Roy will conclude with emerging challenges and opportunities in building compact, trustworthy, and context-aware acoustic systems for the next generation of interactive, perceptive technologies.
Nirupam Roy (Wed,) studied this question.