World’s First Integrated NPU and ISP Image Sensor ASIC Launched, Targeting Driver Monitoring Applications

Leading digital imaging solution developer OmniVision Technologies recently announced the OAX8000 ahead of CES. This automotive-specific application-specific integrated circuit (ASIC) optimized for entry-level Driver Monitoring Systems (DMS) utilizes AI technology. The OAX8000 features a chip-stacking architecture, providing the industry’s only DMS processor with on-chip DDR3 SDRAM memory (1GB). It is also the only dedicated DMS processor that integrates a Neural Processing Unit (NPU) and an Image Signal Processor (ISP), offering dedicated processing speeds of up to 1.1 trillion operations per second for gaze and eye-tracking algorithms. Thanks to its high processing speed, 1K MAC convolutional neural network (CNN) acceleration, and integrated SDRAM, the DMS system can achieve low power consumption—both the OAX8000 and OmniVision’s automotive image sensor have a total power consumption of only 1 watt under typical conditions. This integration further optimizes the DMS system and reduces the circuit board area of the Engine Control Unit (ECU).

According to Yole, with the acceleration driven by the market, the compound annual growth rate (CAGR) for DMS is expected to reach 56% from 2020 to 2025. Under the Euro NCAP requirements of the European Union, all new cars sold in the region must be equipped with DMS cameras by 2022, which is advancing the development of DMS.

“Most DMS processors currently on the market are not specifically designed for this application, requiring additional circuitry to run other system functions, which increases power consumption, occupies more circuit board space, and does not provide space for on-chip SDRAM,” said Brian Pluckebaum, Automotive Product Marketing Manager at OmniVision Technologies. “We focused the design of the OAX8000 ASIC on entry-level DMS, developing the most optimized solution for the automotive industry.”

The on-chip NPU of the OAX8000 supports widely used toolchains such as TensorFlow, Caffe, MXNet, and ONNX. Additionally, this ASIC embeds a quad-core QuadArm® Cortex® A5 CPU with Neon™ technology to enhance the speed of video encoding/decoding and on-chip video analysis algorithms, along with hardware for image processing, video encoding, and RGB/IR processing. With high dynamic range (HDR) processing capabilities, the ASIC can accept inputs from RGB/IR image sensors and support high-quality output, suitable for capturing video in conditions with strong light contrast during both day and night. The integrated video encoder can receive images captured by OmniVision’s automotive image sensors of up to 5 million pixels and output video at resolutions up to 2K at 30 frames per second.

World's First Integrated NPU and ISP Image Sensor ASIC Launched, Targeting Driver Monitoring Applications

The OAX8000 has a significantly shorter startup time compared to its closest competitors. This rapid startup eliminates any delay between the ignition and activation of the DMS camera. Furthermore, the OAX8000 supports secure boot functionality, providing cybersecurity. Other applications include processing occupant detection algorithms, such as distinguishing between infants and grocery bags, and issuing alerts when objects are left in the vehicle. Additionally, this ASIC can be used in automotive video safety systems to run functions such as FaceID and activate preset driver comfort settings (e.g., seat position) when the DMS first scans the driver’s face. The newly launched OAX8000 ASIC is available in BGA196 packaging. This product has passed automotive application AEC-Q100 Grade 2 certification.

Further Reading: “Time-of-Flight (ToF) Sensor Technology and Applications – 2020 Edition” “LiDAR Industry and Core Components – 2020 Edition”

“Autonomous Driving Technology and Market – 2020 Edition”

World's First Integrated NPU and ISP Image Sensor ASIC Launched, Targeting Driver Monitoring Applications

Leave a Comment