Synaptics Inc. has launched an Edge AI evaluation kit (EVK) to help designers develop and prototype artificial-intelligence applications at the edge. Based ON Synaptics’ low-power Katana system-on-chip (SoC) platform, the kit combines vision-, motion-, and sound-detection hardware and software with wired and wireless connectivity.
At the heart of the AI evaluation kit is the low-power Katana SoC with its advanced neural network engine, in combination with multiple sensors, software, and wireless connectivity. The kit targets edge AI for the IoT, including smart-home, building, industrial, and monitoring applications. These include applications for smart-home security and assisted living, industrial vision solutions, smart buildings, and smart cities.
The full-featured evaluation kit builds on Synaptics’ three-pronged strategy approach to become a strong player in the internet of things. The three cornerstones of the strategy is to augment its technologies and products around wireless sensing, edge AI, and low-power processing at the edge. Over the past several years, the company has gained expertise in these areas through acquisitions like Broadcom’s wireless IoT business, Conexant Systems LLC, Marvell Technology Group’s Multimedia Business Unit, DisplayLink Corp., and, most recently, DSP Group.
Click for a larger image. (Image: Synaptics)
Historically, Synaptics has largely focused on PC peripherals and touch interfaces and it is still very important, but the company needed new avenues of growth, said Ananda Roy, senior product manager for Low-Power AI at Synaptics. Edge AI has lots of room to grow and the executive team decided to pivot the company with the cash flow from the traditional PC business into IoT, AI, and edge-AI–centric business models, he said.
“This is where we are going to grow over the next five to 10 years and what all these initiatives on low-power AI, edge AI, and tiny machine learning have been all about,” said Roy. “With the acquisition of the DSP Group, which has an incredibly strong portfolio in the low-power AI space, especially on the audio and the voice side of things, this comes together for our overall AI and IoT strategy.”
These acquisitions, coupled with the gains in technology expertise, in addition to its interface products, have made Synaptics “a one-stop shop not only for touch and interface devices, but also now for low-power AI and premium sound and audio features,” said Roy. “We can offer more to the end customer, and that is an important aspect of what we are doing.”
Synaptics: A natural progression to edge-AI processors
What’s common in the applications for low-power edge AI solutions is “they all require intelligent and efficient detection and analysis of image, audio, voice, and motion data,” he said.
“What has typically driven IoT has been the advances in connectivity technology like 5G, LTE, and all the different generations of Wi-Fi, so it was all about connecting devices to the internet,” said Roy. “The sensing data would be collected and then sent over to the cloud, where a lot of decision-making would happen.”
But over time, it became apparent that it wasn’t the most efficient way to make decisions, so why not do as much as possible at the edge where the data is collected, which frees up expensive data bandwidth and improves privacy and security issues, he added.
In some instances, various sensor fusion algorithms are required to provide more accurate and actionable data, Roy said. “Edge AI and sensor fusion are complex, multifaceted fields of application development, and we know from experience that our customers benefit most by having a more complete platform of hardware, software, wireless connectivity, and technical support to get their IoT concepts to market.”
Historically, or until now, there have been two types of kits coming from Semiconductor companies, said Roy. One is a full-blown development kit, which packs a lot of things and is usually in a bigger form factor, and on the other side of the spectrum is a small evaluation kit for a certain type of application or applications and it doesn’t have many developmental capabilities, he added.
With the Katana Edge AI EVK, it is a small form factor packed with features like vision and audio processing on-chip, with a dedicated neural network accelerator, and it gives engineers the ability to develop low-power vision AI solutions without needing any companion motherboards, he said.
The Katana Edge AI EVK includes a baseboard with the Katana SoC with advanced neural network engine for low-power edge AI processing, Synaptics’ Wi-Fi + Bluetooth Low Energy wireless connectivity module, and sensors for image, audio, voice, and motion.
The Katana Edge AI SoC platform features the company’s proprietary neural network and domain-specific processing cores, on-chip memory, and use of multiple architectural techniques for power savings. It is built on a multi-core processor architecture optimized for ultra-low power and low latency for voice, audio, and vision applications. And as highlighted in the new EVK, it can be combined with the company’s wireless connectivity offerings for system-level solutions.
The EVK also includes accessories for people-detection application development (such as mounting bar and clips for overhead camera), cabling, batteries, and enclosure, along with the instructions and software. In addition, there are associated tools to enable optimization of the AI algorithms.
The Edge AI EVK is available now. Two demos are available: the Katana Edge AI EVK reading meter display panel and the Katana Edge AI EVK performing people detection and counting.