A processing unit that integrates signal processing and artificial intelligence (AI) algorithms into Micro-Electro-Mechanical Systems (MEMS) sensors to support local decision-making while significantly reducing space and power consumption is expected to realize the convergence of electronic decision-making technology and the physical world.

STMicroelectronics (ST) announced on the 21st that it is launching an Intelligent Sensor Processing Unit (ISPU) that integrates a DSP (Digital Signal Processor) and MEMS sensor suitable for executing AI algorithms on the same silicon.

This ISPU is a System-in-Package (SiP) device that not only reduces size and power consumption by up to 80%, but also integrates sensors and AI to enable electronic decision-making at the application edge. This will accelerate the Onlife Era, where innovative products enabled by smart sensors will be able to sense, process, and act, realizing the convergence of technology and the physical world.

The Onlife era refers to everyday life where connected technologies continuously support natural and transparent interactions and seamless transitions between online and offline, making it impossible to distinguish between online and offline. ST is realizing the Onlife era by enabling the intelligent processing that supports this lifestyle to be transferred to sensors at the edge through the ISPU.

ST's ISPU offers significant advantages in four areas: power consumption, packaging, performance, and cost. The unique ultra-low-power DSP is programmable in the C language, a language familiar to many engineers. Quantized AI sensors can support neural networks with single-bit precision, as well as full-bit precision. This ensures exceptional accuracy and efficiency in tasks such as activity recognition and anomaly detection through inertial data analysis.

“Integrating ST sensors and ISPUs on the same silicon was technically challenging, but the ISPU advances sensor functionality to accelerate decision-making by reducing data transmission, enhance privacy by keeping data local, and reduce size, power, and cost, enabling sensor-based systems to elevate from online to on-life experiences,” said Andrea Onetti, Executive Vice President, MEMS Subgroup, STMicroelectronics. “Especially because the ISPU is easily programmable with commercial AI models and will ultimately work with all major AI tools.”