Nanoveu’s EMASS technology achieves real-time, always-on keyword spotting using bone-conduction vibration on the ECS-DoT edge AI chip at under 1 milliwatt power, eliminating the need for microphones and overcoming key challenges in wearable voice interfaces.
- Bone-conduction sensing enables noise-immune, low-power voice detection
- Keyword spotting runs always-on below 1 milliwatt on ECS-DoT AI SoC
- Privacy enhanced by excluding ambient audio from signal path
- Applications span hearables, AR/VR, industrial and defence sectors
- Active commercial engagements progressing towards production-ready designs
Breakthrough in Wearable Voice Sensing Technology
Nanoveu Limited (ASX:NVU) has pushed the boundaries of wearable voice interfaces by demonstrating a real-time keyword spotting system that operates without a microphone and consumes less than 1 milliwatt of power. This breakthrough, achieved through its EMASS technology running on the ECS-DoT edge AI system-on-chip, uses bone-conduction sensing to capture speech vibrations directly from the jawbone rather than air. The result is a voice interface that simultaneously solves the long-standing trade-offs between noise immunity, power consumption, and latency.
Traditional voice-activated devices struggle to balance these three factors: microphones pick up ambient noise requiring power-hungry cancellation, always-on listening drains battery life, and motion-triggered wake systems add latency. Nanoveu’s approach sidesteps these compromises by sensing micro-vibrations through bone conduction, inherently excluding environmental noise and eliminating the need for complex noise-cancellation processing.
Sub-Milliwatt Always-On Operation and Enhanced Privacy
The ECS-DoT AI chip, independently benchmarked at up to 20 times the energy efficiency of leading edge-AI chips, enables the entire keyword spotting application to run continuously at under 1 milliwatt. This ultra-low power envelope means devices can maintain always-on voice detection without significant battery drain, a critical factor for compact, battery-constrained wearables like earbuds and AR glasses.
Moreover, by never capturing ambient audio, the system inherently protects user privacy, reducing the risk of inadvertent recording of nearby conversations. This privacy advantage comes without additional hardware or processing overhead, distinguishing Nanoveu’s solution in a market increasingly sensitive to data security concerns.
Market Potential Across Hearables and Beyond
The global hearables market is projected to grow from USD 62.22 billion in 2026 to USD 107.1 billion by 2031, driven by consumer demand for smarter, voice-controlled devices. Nanoveu’s noise-immune, low-power voice sensing technology aligns perfectly with this trend, enabling natural, hands-free interaction in challenging acoustic environments such as factories, busy streets, and public transport.
Beyond hearables, the bone-conduction sensing platform has applications in AR glasses, VR headsets, professional and industrial communication equipment, and defence systems where reliable, discreet, and battery-efficient voice control is essential. The company is actively engaging with device manufacturers and module makers to develop production-ready reference designs tailored for these markets.
Technical Innovation and Commercial Progress
The demonstration employed a standard STMicroelectronics inertial measurement unit (IMU) to detect jawbone vibrations, paired with the ECS-DoT AI chip; no specialised hardware was needed. This underscores the flexibility and scalability of Nanoveu’s approach. Achieving sub-milliwatt inference required aggressive AI model compression and scaling, alongside advanced on-chip energy management, showcasing ECS-DoT’s capability to perform edge AI inference on vibration-based sensing.
Nanoveu’s CEO of Semiconductor Technologies, Mark Goranson, emphasised the broader implications: “Recognising speech directly from bone-conduction vibration at under one milliwatt, with no microphone in the signal path, shows the chip can extract a meaningful signal from mechanical vibration in real time, within a power budget that compact, battery-powered devices can support.” This positions ECS-DoT not only as a hearables enabler but as a versatile platform for diverse vibration-sensing applications.
This milestone builds on Nanoveu’s recent progress, including the tape-out of its 16nm ECS-DoT chip and expansion into drone systems, reflecting a multi-pronged strategy to commercialise ultra-low-power edge AI across sectors. The company is advancing keyword vocabulary expansion, detection robustness, and production-ready designs while pursuing design-in engagements with ecosystem partners.
Bottom Line?
Nanoveu’s microphone-free, ultra-low-power voice sensing could redefine wearable AI interfaces, but commercial adoption will hinge on scaling detection capabilities and securing design wins.
Questions in the middle?
- How quickly will Nanoveu expand keyword recognition beyond current capabilities?
- What timelines are realistic for integration of ECS-DoT in consumer hearables and AR devices?
- Can Nanoveu maintain its energy efficiency advantage as competitors develop similar edge AI solutions?