Syntiant Releases TWS Earbud Reference Design Based on NDP100 Neural Decision Processor

By Abhishek Jadhav

Freelance Tech Writer

February 16, 2022


Syntiant Releases TWS Earbud Reference Design Based on NDP100 Neural Decision Processor

Syntiant introduced a reference design for true wireless earbuds to enable faster time to market for ODMs and OEMs. The ultra-low-power design comes with the NDP100 neural decision processor that provides more than 100x performance and 10x throughput at under 140 microwatts for always-on AI voice processing.

An important aspect of the design lies in its voice assistant system. Users are looking for a hands-free, voice-enabled AI ecosystem to control device functionalities. The design supports multiple languages including Mandarin and Cantonese, increasing the market scope. The ability to freely control music through local speech commands comes from the edge AI reference design, which incorporates a highly accurate voice interface for TWS earbuds, increasing efficiency while consuming minimal power.

The hardware reference design also features edge-optimized training pipelines that open the door to building neural networks for a wide range of always-on voice, sensor, and image applications. The flexibility brought in by the reference design gives ODMs the opportunity to customize the neural networks for application-specific tasks. Deep learning processing capabilities in the TWS earbud edge device reduce dependency on cloud processing while ensuring privacy and security.

(Image Credit: Syntiant)

The introduction of the NDP100 is based on a foundation of deep learning algorithms featuring an embedded Arm Cortex-M0 processor with integrated 112kB SRAM. The device features tight integration of several AI subsystems to achieve neural network processing with over half a million parameters for sensor-based applications. The NDP100’s programmable Core 1 deep neural network supports application-defined audio sequences for a variety of applications. This became the highlighting point for the reference design based on the keyword speech interface, wake word detection, speaker identification, and audio environment classification methodologies.

The voice-enabled reference design enables data and training support and multilingual voice compatibility while reducing the development cost. ODMs and OEMs across the globe can use Syntiant’s reference design to develop and deploy ultra-low-power, edge AI voice processing applications in true wireless earbuds.

Abhishek Jadhav is an engineering student, freelance tech writer, RISC-V Ambassador, and leader of the Open Hardware Developer Community.

More from Abhishek