ECS-DoT: A Milliwatt-Class Leap for Always-On AI
EMASS’s ECS-DoT is an ultra-low power voice detection chip built as a full AI system-on-chip, engineered for continuous, on-device intelligence. Announced ahead of its showcase at Sensors Converge, the ECS-DoT targets a central challenge in edge AI wearables: keeping features always on without sacrificing battery life. Positioned as a breakthrough low power SoC, EMASS claims the chip can deliver 10–100× lower energy consumption and up to 3× faster AI inference for common tasks compared with existing solutions. This combination of performance and efficiency is designed to support ultra-low power voice detection, keyword spotting, and sensor fusion while remaining within a milliwatt-class power envelope. By pushing AI processing directly to the device, ECS-DoT aims to eliminate the traditional trade-off between responsiveness and power drain, laying the groundwork for truly always-on AI chips in compact, battery-powered products.
Solving the Voice Recognition Battery Problem in Wearables
Always-on voice assistants have long been constrained by a simple reality: microphones and AI accelerators consume energy, quickly eroding battery life in small devices. EMASS positions ECS-DoT as a direct answer to this voice recognition battery problem. Its RISC-V architecture and non-volatile memory stack enable AI workloads to run at very low power, allowing continuous listening and detection without the need for frequent charging or oversized batteries. For edge AI wearables like smart glasses, hearables, and hearing aids, this means wake-word detection and ambient listening can stay active in the background without overwhelming energy budgets. Instead of offloading data to the cloud or relying on high-power CPUs, ECS-DoT handles tasks such as keyword spotting locally, reducing latency and preserving privacy. The result is an always-on AI chip that supports natural voice interaction while respecting the tight power and space constraints of modern wearables.
Bone-Conduction Voice Sensing: A New Mode of Private Interaction
One of the most intriguing demonstrations of ECS-DoT focuses on smart glasses using bone-conduction voice detection. EMASS integrates an Inertial Measurement Unit directly into the frame, enabling the chip to sense subtle jaw vibrations transmitted through the temple arm. Rather than leaving microphones permanently active, ECS-DoT analyzes these vibrations to perform voice activity detection and keyword spotting. For users, this approach promises more private, low power voice interaction, since the system can detect speech without constantly capturing ambient audio. For designers of edge AI wearables, it reduces reliance on traditional always-on microphones, freeing up power and helping minimize acoustic leakage. This sensor-fusion strategy illustrates how a low power SoC can unify motion and audio signals to unlock new forms of interaction, moving beyond conventional voice interfaces toward more discreet, context-aware control in everyday eyewear and hearables.
From Smart Glasses to Drones: A Platform for Edge AI Wearables and Beyond
While smart glasses are the headline application, EMASS positions the ECS-DoT as a general-purpose platform for edge AI wearables and other battery-constrained systems. The chip is optimized for highly compressed AI models and can handle real-time processing of audio, vision, and motion data. That versatility extends its potential into drones, industrial sensing, and autonomous machines, where extending flight time or operating duration is as critical as response speed. EMASS highlights drone endurance and on-device processing gains as early performance indicators, aligning with broader industry efforts to shift intelligence from cloud to edge. A flexible SDK and a development-to-production pipeline aim to help manufacturers quickly deploy custom models on the ECS-DoT. By combining low latency inference, integrated sensor fusion, and ultra-low power voice detection in a single always-on AI chip, EMASS is signaling a new design template for next-generation edge devices.