Syntiant, GlobalSense introduce edge AI solution for automotive sector
Syntiant Corporation, an edge AI deployment company, and GlobalSense, a developer of sound-based machine learning applications for the auto industry, have jointly unveiled two deep learning solutions. These cloud-free solutions are designed to enhance vehicle security and safety.
Utilizing Syntiant’s TinyML board with embedded NDP101 neural decision processor and GlobalSense’s AI models, the companies note that the low-power applications enable audio event detection (AED) and predictive maintenance.
According to Kurt Busch, CEO of Syntiant, its neural decision processors are ideal for battery-powered applications that require advanced deep learning processing. He adds that the combined hardware and software technologies provide accurate, always-on event detection, eliminating noise and false detects.
“Whether it’s ensuring occupant safety while driving or detecting break-ins and vandalism when the vehicle is parked, all of this is achieved without relying on the cloud, with minimal power consumption, and at significantly lower cost,” Busch notes.
The applications can also recognize and respond to specific sounds like tires screeching and glass breaking. Instant notifications are sent to predefined recipients via text or email, including the vehicle’s owner, security service providers or the appropriate authorities, company executives say.
Saeid Safavi, Ph.D., co-founder and CEO of GlobalSense, emphasizes the neural platform provided by Syntiant’s processor technology for their endpoint AI. He also highlights its cost-effectiveness and energy efficiency.
The Syntiant NDP101 also executes audio and sensor applications using deep learning algorithms. Company executives say it operates with a power consumption of less than 140 microwatts.
Syntiant, at the beginning of 2023, introduced the NDP115 chip, a neural decision processor designed for various applications, including home security and industrial IoT. The chip utilizes at-memory compute and consumes less than one milliwatt of power while running multiple neural network operations, according to the company.
Edge World Summit 2024
Comments