Femtosense, in partnership with ABOV Semiconductor, recently unveiled the AI-ADAM-100, an artificial intelligence microcontroller unit (AI MCU) utilising sparse AI technology. This innovation enables on-device AI features, such as voice-based control in home appliances and other products, providing immediate, low-latency user responses with reduced power consumption, enhanced security, operational stability, and cost-effectiveness compared to GPUs or cloud-based AI.
The AI-ADAM-100 combines the Femtosense Sparse Processing Unit 001 (SPU-001), a neural processing unit (NPU), with an ABOV Semiconductor MCU, offering deep learning-powered AI voice processing and voice-cleanup capabilities on-device at the Edge. This integration allows appliances to implement natural voice interfaces, enabling users to speak freely and be understood in multiple ways. For instance, phrases like “Turn the lights off,” “Turn off the lights,” and “Lights off” all convey the same intent and are recognised as such. The voice/audio cleanup process enhances data reliability and accuracy before transmission to the cloud, reducing the volume of data sent and, consequently, backend infrastructure costs.
“With sparsity integrated throughout the AI development stack, the AI-ADAM-100 is the first device on the market to fully unlock the advantages of sparse AI,” explained Sam Fok, CEO, Femtosense. “Our sparsity-enabling technology allows our customers to deliver compact, efficient AI processing to a growing variety of markets and products, including home appliances as well as small form factor, battery-operated devices like high-fidelity hearing aids, industrial headsets, and consumer earbuds.”
In addition to the AI-ADAM-100, Femtosense offers a versatile selection of AI-ADAM-100–based AI software applications—from turnkey solutions to tool-driven applications or bespoke implementations using a manufacturer’s own AI models, whether dense or sparse.
Sparse AI optimises the cost of AI inferencing by eliminating irrelevant portions of an algorithm and allocating hardware memory and compute resources only to the remaining relevant parts. A system that stores and processes only nonzero weights can achieve up to a 10x improvement in speed, efficiency, and memory footprint.
Similarly, a system that processes only when a neuron’s output is nonzero can realise up to another 10x increase in speed and efficiency. These multipliers enable sparse AI to implement deep learning models up to 100 times more powerful or complex than previous MCUs without negatively impacting speed, efficiency, memory footprint, or performance.
While many Edge applications could benefit from AI, they often lack the financial or power flexibility to deploy a GPU, maintain cloud connectivity, or support a dedicated silicon solution, limiting the adoption of Edge AI. The AI-ADAM-100 allows manufacturers to implement voice language interfaces at the Edge, even for devices not connected to the cloud.
Most current AI systems continuously process and consume power, even during simple tasks like quiet environments. Pure, Cloud-based voice processing requires continuous throughput, leading to high infrastructure costs. The AI-ADAM-100 addresses tasks on-device, significantly reducing power consumption and backend cloud load. Specifically, it enables home appliance manufacturers to implement sophisticated wake-up and control functionality, allowing other system controllers and connectivity modules to enter sleep mode and consume less power when not in use. This capability allows the system to listen for user commands and either process them on-device or wake the system to send the command to the cloud.
The result of a partnership
The AI-ADAM-100 MCU is a result of strategic collaboration between Femtosense and ABOV, leveraging the core strengths of each partner. “The AI-ADAM-100 is the best-optimised AI MCU solution for voice and audio-based AI applications and enables a variety of on-device AI applications for consumer electronics and standalone devices,” said Choi Won, CEO of ABOV Semiconductor. “Together with Femtosense, we will continue to develop the most cost- and power-efficient AI MCUs for global customers.”
ABOV has verified the AI-ADAM-100’s superior voice command recognition performance under various noise conditions, meeting the requirements of leading customers. Global home appliance manufacturers are working to reduce the number of buttons on their devices and streamline the user experience. AI-based voice command can significantly accelerate this trend.
There’s plenty of other editorial on our sister site, Electronic Specifier! Or you can always join in the conversation by commenting below or visiting our LinkedIn page.