Ultra-low-power sensor board targets machine learning at the edge

By Gina Roos, editor-in-chief

Eta Compute Inc. has claimed the industry’s first integrated, ultra-low-power AI sensor board, designed for machine learning at the edge. The ECM3532 AI sensor board reduces the initial phase of smart sensor development — including feasibility, proof of concept, and board design — from several months to weeks or even days by using the company’s TENSAI SoC , according to the company.

Previously, smart sensor design required developers to use multiple boards together. With the integrated ECM3532 board, nearly all the functionality required, including data acquisition, is integrated into a single board that works for months on a single battery cell while performing machine-learning algorithms. The board includes two pulse-density-modulation (PDM) microphones, one pressure and temperature sensor, and one six-axis MEMS accelerometer and gyroscope.

In addition, the small form factor, measuring 1.4 × 1.4 in., along with the embedded battery socket and Bluetooth connectivity with the ABOV Semiconductor A31R118 BLE SoC, makes it suited for IoT deployment and field testing application prototypes. Applications include voice activation and voice commands in consumer, home, and building automation products; asset tracking and monitoring and predictive maintenance in industrial products; and context awareness in consumer and industrial applications. The board can be purchased through Digi-Key.

The company is hosting a free online workshop with Edge Impulseon July 14, 2020. The first 50 qualified developers planning or working on projects will receive a complimentary ECM3532 AI sensor board. You can register for the workshop here .