Femtosense introduces the AI-ADAM-100 system-in-package for cost-effective, efficient AI-based language processing at the edge

The latest version of Femtosense, the AI-ADAM-100, combines a low-power Arm Cortex M0+ microcontroller from ABOV Semiconductor with the Femtosense Sparse Processing Unit (SPU), an in-house neural processing unit. This combination enables AI language processing and sanitization capabilities on edge devices.

It leverages sparsity techniques to ensure that models can run on edge devices without intensive processing. Sparsity in AI refers to the presence of zeros or non-zero values ​​in the matrices and tensors used in machine learning models. No computation is required for these zero weights, which significantly reduces the processing overhead.

The Femtosense AI-ADAM-100 is designed to enable voice control for various devices such as home appliances, hearing aids, industrial headsets and consumer earphones. These devices can implement voice user interfaces that allow users to speak naturally in their own words. An on-device voice sanitization feature would reduce infrastructure costs and improve the accuracy of data sent to the cloud. The Femtosense SiP also allows home appliance manufacturers to implement wake words to reduce power consumption.

Femtosense AI-ADAM-100 System-in-Package Specifications:

• MCU – ABOV Semiconductor ARM Cortex-M0+ microcontroller
  • Flash Memory – 64/32 KB MCU code flash memory
  • SRAM – 8KB
• Co-processor – Femtosense SPU-001 NPU with 1 MB NPU SRAM (10 MB effective with sparsity)
• NPU – 22 nm process | MCU: 130 nm process
• Sub-mW receiver for voice, audio and 1-D data
• Host Interface – SPI

Femtosense provides a Software Development Kit with advanced sparse optimization tools, a hardware simulator, and the Femtocrux compiler for deployment. Custom AI models can be deployed with or without fine-tuning in the sparse processing unit. Femtosense also provides some ready-to-use demo AI models for noise cancellation and speech interfaces.

There are currently no development boards available for testing the AI-ADAM-100, but there are two evaluation boards for the SPU-001 neural processing unit. The evaluation board (EVB) supports integration with external host systems with 16MB of onboard flash and a PMOD header that provides full access to the SPU pins. The evaluation kit 2 (EVK2) is more powerful and features a Cortex-M7 host processor (with floating point unit), 2x onboard MEMS microphones, 3.5mm audio jacks for input/output and a microSD card reader. It also integrates a host connector board from the open source hearing aid platform Tympan.

We have covered other edge AI products and MCU-grade chips, such as the Infineon CY8CKIT-062S2-AI evaluation kit, the Ambiq Apollo510 Arm Cortex-M55 MCU, the Tokay Lite ESP32-S3 no-code AI camera, and more.

The AI-ADAM-100 is currently in the pre-production phase. Engineering samples can be requested from the company. Mass production is expected to begin later this year. For more information, see the product page and press release.

Thanks to TLS for the tip.

Tomisin is a writer specializing in hardware product reviews, comparisons, and explanations. He is passionate about small form factor and single-board computers.

Support CNX Software! Donate with cryptocurrencies, become a patron on Patreon or buy goods on Amazon or Aliexpress.