EMASS researchers at KAIST demonstrated a programmable neuron that achieved 94.8% accuracy in human activity recognition and 95.0% in speech recognition using semiconductor tuning.
Public source
Publisher name
Public post
Voice and motion don't move at the same speed. Speech signals reach into the kHz range; everyday motion sits in the Hz range — different signals entirely. Researchers at…
Company
EMASS
- Industry
- Semiconductor Manufacturing
- Company size
- 11–50 employees
About EMASS
EMASS, a Nanoveu business, is redefining what’s possible in ultra-low-power Edge AI. Our first product, the ECS-DoT, is a breakthrough SoC that delivers 10–100× lower energy consumption and up to 3× faster inference for common AI tasks compared to existing solutions. Optimized for highly compressed AI models and featuring integrated sensor fusion, ECS-DoT enables real-time processing of audio, vision, and motion data in compact, battery-powered devices. Whether extending flight time in drones, powering always-on features in AR glasses, or enabling predictive maintenance in industrial IoT, EMASS brings intelligence to the edge—where size, power, and latency matter most. With a flexible SDK that supports multiple model types and a proven development-to-production path, EMASS gives customers the fastest route to differentiation and design wins in next-gen edge devices.
See moreLatest activity
Latest activity from EMASS
3 signals
Presence & Recognition
EMASS Field Application Engineer Brendan Schuster will present Event-Driven SoC Design for Ultra-Low-Power Edge AI at Embedded World North America on Tuesday, September 22 at 4:20 PM (Day 1) at the Anaheim Convention Center.
Products & Services
EMASS featured Scott Smyser on a recent episode of PSDcast discussing the Self-Aware Asset Tag collaboration with Bosch Sensortec.
Discover more
Similar signals
Similar public activity from other companies.
Research & Knowledge
Hprobe
Hprobe published a new study demonstrating 300 mm-compatible SOT-MRAM and evaluating its use in analog crossbar applications including neural-network inference, stochastic training, and probabilistic graph modeling.
Research & Knowledge
Samsung Semiconductor
Samsung Semiconductor is engineering memory technologies to meet the demands of Edge AI, including capacity, bandwidth, and energy efficiency.
Research & Knowledge
Analog Devices
Analog Devices highlights Jerry Tworek's venture on test-time learning, arguing that agents learn while doing and retaining experience without requiring a fine-tuning cycle.
Research & Knowledge
Kopin Corporation
Kopin Corporation published a white paper exploring the modeled opportunity in Neural I/o™ and AI data movement.
Research & Knowledge
AMD