Arduino Puts a Syntiant NDP Machine Learning Chip on Its New Nicla Voice TinyML Development Board
Featuring a Syntiant NDP120 ultra-low-power accelerator, this new dev board looks to bring on-device voice processing to your next project.
UPDATE (2/2/2023): The Arduino Nicla Voice board, an Arduino Pro device which includes a Neural Decision Processor (NDP) from edge AI specialist Syntiant, is now available to purchase.
"Small enough to fit into wearables or retrofit existing machinery, enabling AI yet requiring minimal energy," the Arduino team claims of the board, unveiled early last month, "Nicla Voice is the 'impossible' combination that makes voice recognition on the edge possible – and easier than ever."
The Nicla Voice is selling on the Arduino Store for $82, making it the cheapest model in the growing Nicla range yet.
Original article continues below.
The Arduino team has officially launched the next entry in the compact Nicla development board family, this time focusing on tinyML voice processing at the edge with an integrated Syntiant Neural Decision Processor (NDP): the Arduino Nicla Voice.
"With Nicla Voice’s ready-to-use combination of sensors and processing power,," the Arduino team says of its latest launch, "you can prototype and develop new solutions that leverage voice detection and voice recognition, or interpret any other audio input – from machines that need maintenance to water dripping, and from glass breaking to alarms that must get through headphones' noise-canceling features. We can’t wait to hear what you’ll create with it!"
Built in the smallest official Arduino form factor, which launched with the Bosch-partnered Nicla Sense ME in September 2021 and was followed by the Nicla Vision in March last year, the Arduino Nicla Voice is built around a Nordic nRF52832 system-on-chip with a single Arm Cortex-M4 microcontroller core running at up to 64MHz, 64kB of static RAM (SRAM), 512kB of on-chip flash plus 16MB SPI flash for additional storage, and a Bluetooth Low Energy (BLE) radio for wireless connectivity.
The board also includes integrated sensors, starting with — unsurprisingly, given the device's focus on edge-AI voice processing — an Infineon XENSIV IM69D130 MEMS microphone. Elsewhere is a Bosch BMI270 ultra-low-power six-axis inertial measurement unit (IMU) with integrated gesture, context, and activity recognition plus step counting and a Bosch BMM150 three-axis magnetometer.
It's the board's coprocessor that's of the most interest, however: the Nicla Voice is the first Arduino to come equipped with a Neural Decision Processor (NDP) from Syntiant, designed to provide high-performance on-device machine learning without high power draw. The development board uses Syntiant's NDP120, which launched in early 2021 as the company's first to be based on the second-generation Syntiant Core 2 tensor processor architecture — and which delivers, the company claims, a twenty-fivefold performance boost over its first-generation decision processor parts.
The NDP120 includes support for up to seven audio streams, I2S and TDM audio output support, its own dedicated Arm Cortex-M0 microcontroller, and 48kB of dedicated SRAM — entirely separate to the 64kB SRAM available on the board's primary microcontroller. "We took years of real world, low-power edge deep learning experience to develop this architecture into a scalable design optimized to bring neural processing to power constrained deployments," Kurt Busch, Syntiant's chief executive, claimed at the time.
As with previous Nicla boards, the Arduino Nicla Voice features castellated pins for breadboard or surface-mount installation, with a total of two analog to digital converter (ADC) pins, an SPI bus, an I2C bus brought out to a five-pin ESLOV connector, and a single UART. A ZIF connector allows for the use of an external microphone, while the input/output voltage is switchable in software from 1.8V to 3.3V.
More information on the Arduino Nicla Voice is available on the Arduino Pro site. The board is listed on the Arduino Store for $82, with interested parties asked to subscribe to a waiting list to place their order.