STMicro's New iNEMO Chip Hands Six-Axis IMU, Bone-Conduction Audio, and Machine Learning Tasks
Designed for "hearables" and VR/AR headsets, STMicroelectronics' tiny chip is an all-in-one tool with impressive potential.
STMicroelectronics has announced an all-in-one chip which includes a six-axis inertial measurement unit (IMU) and a bone-conduction system for voice detection — aiming at compact wearables and "hearables" including wireless headphones and virtual or augmented reality headsets.
STMicro's new iNEMO LSM6DSV16BX system-in-package (SIP) for wearables is based around a triple-core design, which the company says enhances efficiency. A Finite State Machine (FSM) provides gesture recognition support without straining the primary processor, a Machine Learning Core (MLC) handles activity recognition and voice detection on-device, and an Adaptive Self-Configuration (ASC) system keeps overall power draw low while offloading tasks from the host processor.
That voice detection, however, isn't driven by a microphone. Instead, the chip offers a dedicated audio accelerometer which picks up the wearer's speech through bone conduction — effectively doing away with the problem of background noise.
For head tracking, activity detection, and 3D-sound tasks, meanwhile, the chip includes STMicro's Sensor Fusion Low Power (SFLP) system with a six-axis inertial measurement unit (IMU) built-in — and dedicates a third channel, alongside motion tracking and audio capture, for quasi-electrostatic potential (Qvar) sensing for touch-based interaction.
According to STMicro's reckoning, using the LSM6DSV16BX can cut overall power consumption by 70 per cent and PCB area by 45 per cent compared to separate alternatives — and the package comes in some 14 per cent thinner than STMicro's earlier MEMS-based intertial sensor equivalents. The company has also measured the chip's power draw at 0.6mA with the motion tracking system in high-performance mode.
STMicroelectronics has begun production of the chip now, and has made software samples for the FSM and MSC cores available on GitHub, with more information available on the company's website. Pricing has been set at $3.95 per chip in 1,000-unit quantities — with a free sample available to interested parties.