POLYN's Upcoming NASP Neuromorphic TinyML Chips Get Voice Extraction Capabilities with NeuroVoice

Drawing just 100µW of power and small enough for in-ear earbud use, this tinyML chip family can pull clear speech from the noisiest feeds.

Neuromorphic processing specialist POLYN Technology has announced availability of NeuroVoice, a solution for on-chip voice extraction capable of working even in the presence of "any noisy background" audio — designed for use on its ultra-compact neuromorphic processor chips.

"Earbuds, smartphones, hearing assistance, gamer headphones, and intercom systems need new technology to bring voice processing to a new level," claims POLYN's Eugene Zetserov. "Current methods of voice signal processing are power hungry and, in some cases, fall short. Immediate voice recognition is important for hearing assistance devices."

NeuroVoice aims to bring AI-based voice extraction to long-life "hearables," using neuromorphic processing units. (📹: POLYN Technology)

"AI-based extraction of the voice signal in a noisy environment, including irregular noises, provides a better hearing experience than standard noise cancellation filters," Zetserov continues. "A neural network is the perfect tool for voice processing and POLYN offers it on a tiny neuromorphic analog chip."

NeuroVoice, the company explains, takes advantage of the Neuromorphic Analog Signal Processing (NASP) architecture found on POLYN's brain-mimicking edge-AI processors to offer on-chip voice detection and voice extraction from "any background conditions" in a supremely low power envelope: POLYN claims NeuroVoice draws just 100µW in operation and finishes inference in just 20µs.

The company is targeting the "hearables" market with NeuroVoice, including earbuds, hearing aids, gaming headphones, and intercoms, as well as voice-activated assistance systems in the smart home and industry — and has even proposed integrating the technology into future smartphones to allow for improved noise-free audio calls. In addition to voice detection and extraction, the platform is also capable of handling wake-word and keyword detection ― though this functionality won't be available in the first parts to hit market.

Polyn showed off the first test chips based on its NASP technology back in April, proving the concept on a 55nm CMOS process. At the time, it projected mass production of finalized chips in the first quarter of 2023 — including the launch of the first commercial product based on the technology, a sports wearable with integrated photoplethysmography (PPG) heart-rate and inertial measurement unit (IMU) motion sensors.

More information on NeuroVoice is available on POLYN's website; at the time of writing the company had not provided an updated roadmap for the first NASP parts to hit mass production but has confirmed it is accepting orders for NV100 and NV101, offering voice activity detection only and voice activity detection plus voice extraction respectively in chiplet or IP block forms targeting 40-64nm process nodes.

Gareth Halfacree
Freelance journalist, technical author, hacker, tinkerer, erstwhile sysadmin. For hire: freelance@halfacree.co.uk.
Latest articles
Sponsored articles
Related articles
Latest articles
Read more
Related articles