ST’s Human Presence Detection (HPD) solution tracks head and posture with AI and VL53L8CP ToF sensor

Back in 2022, we first saw STMicro launch the VL53L5CP ToF presence detection sensor for the PC market with features like user detection, gesture recognition, and intruder alert. Building upon that, the company has recently released its 5th‑generation Human Presence Detection (HPD) solution, built around the VL53L8CP 8×8 multizone Time‑of‑Flight sensor family with edge AI features offered as a turnkey solution for laptops, PCs, monitors, and accessories.

We have seen and even tested earlier ToF sensors like the VL53L0X and VL53L1X, which used a single-photon avalanche diode (SPAD) for single-zone ranging. After that, we saw ST release the VL53L5, which was a major upgrade with up to 64-zone multi-zone support, enabling features like touch-to-focus and object tracking. Now, in the age of AI, ST has launched a 5th-gen turnkey HPD solution built around the VL53L8CP sensor.

STMicro HPD solution specifications:

• Sensor – ST FlightSense VL53L8CP 5th generation multi-zone direct Time-of-Flight (dToF) sensor
• AI features
  • Presence AI – detects user presence in front of the device
  • HOR AI – head orientation tracking for adaptive screen dimming
  • Posture AI – posture-based wellness detection
  • Hand Posture AI – gesture and hand posture recognition
  • Multi-Person Detection (MPD) – alerts on over-the-shoulder privacy breaches
• Functionalities
  • Adaptive Screen Dimming
  • Walk-Away Lock (WAL)
  • Wake-on-Approach (WOA)
  • Gesture Recognition
  • Posture-based Wellness Monitoring

The full presence detection turnkey solution includes the sensor, firmware, pretrained AI models, and a full SDK for easy integration behind IR-pass filters in laptops, monitors, and accessories. AI inference is handled locally on STM32-based PC sensor hubs, which requires no GPUs or cloud processing. The system uses four lightweight neural networks – Presence, Head Orientation, Posture, and Hand Posture AI, each with fewer than 10,000 weights, trained on real-world movement data and optimized for sub-millisecond inference on 32-bit MCUs. There are also preprocessing and background filtering algorithms that help distinguish humans from objects, with the entire detection loop consuming less than 1 mW in standby.

More information about the product can be found in the announcement. The product page has more details about the company’s ToF sensors but nothing specific to the VL53L8CP sensor just yet.

Thanks to TLS for the tip.

Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress. We also use affiliate links in articles to earn commissions if you make a purchase after clicking on those links.