Syntiant kicks off 2023 with an edge AI chip, security solutions and software for machine vision
Syntiant Corporation, a provider of machine learning and edge AI solutions, has introduced the NDP115 edge AI chip. The company is aiming the chip for use in a variety of applications such as home security and industrial IoT.
This chip is a recent addition to its neural decision processors that use at-memory compute and built-in standard CMOS processes. The Syntiant NDP115 neural decision processor uses the Syntiant Core 2 inference engine, which enables the chip to run multiple neural network operations while consuming less than one milliwatt of power. The company also reports that this new neural decision processor can handle speech inference at a low power consumption of 280 microwatts.
The Syntiant NDP115 can run multiple neural networks with different architectures, like CNN, RNN and fully connected networks. Use cases include close-talk, far-field and keyword speech recognition. The device supports I2C and pulse density modulation, allowing sensor fusion and multi-axis acceleration. Syntiant presented this end-to-end deep learning module at CES 2023.
Syntiant has also integrated its Syntiant Core 2 programmable deep learning architecture into the NDP120 neural decision processor. This architecture has an embedded Arm Cortex-M0 microcontroller, providing up to 13 general-purpose input/output pins and an integrated HiFi-3 DSP, which the user can program.
“The NDP115 offers the multi-modal functionality of our Core 2 inference engine in a compact, cost- and power-efficient solution for ultra-power and size-constrained applications,” said Kurt Busch, the CEO of Syntiant. “Combined with our machine learning software models, the purpose-built NDP115 enables developers to easily deploy full audio and sensor processing solutions that address all kinds of consumer and commercial use cases, from home security to industrial IoT.”
Syntiant introduces edge AI software for computer vision applications; use cases showcased at CES 2023
Syntiant has released pre-trained and customizable deep learning models for computer vision applications. The solution includes data, tools and training, enabling fast deployment across various industries, such as smart homes, personal devices, automotive, governments and industrial environments. Syntiant’s edge AI software empowers enterprises lacking in edge expertise to quickly onboard and create edge-based computer vision applications.
“We are building upon our leadership position in voice and audio by offering scores of off-the-shelf machine learning models for edge-based image and vision applications,” said Busch. “These hardware agnostic models can be easily customized and work on a wide range of SOCs, including our own NDP200, which brings together the best of both worlds into a powerful, compact, highly efficient turnkey solution.”
Syntiant demonstrates edge AI security applications at CES 2023
By leveraging its audio, sensor and vision processing capabilities, Syntiant showcased a myriad of edge AI security applications. Some examples of these applications include a deep learning model for low-power security audio event detection, computer vision models for surveillance and a fresh approach for detecting door tampering. Syntiant believes its pre-configured models and customized solutions can improve edge AI performance on resource-constrained edge devices.
Syntiant demonstrated the Amazon Ring Alarm break sensor, which uses the company’s NDP101 neural decision processor to detect a window break while ignoring other ambient noise. The sensor technology used by Ring can detect the sound of a broken window from 25 feet away, alerting the user instantly via an application notification.
“With the addition of our hardware-agnostic computer vision models alongside our edge-optimized NDP200 and NDP120 silicon, we have developed a complete solution for quickly and seamlessly deploying machine learning applications in security, access control and monitoring,” said Busch.
Article Topics
AI/ML | computer vision | edge AI | neural networks | surveillance | Syntiant
Comments