ToF Sensors for PC Privacy Confirm Even AI Needs Analog Sensors

If you follow the general media and even technical sources, it’s easy to assume that most (but not all) artificial-intelligence (AI) functions consist of high-power specialized processors (in both senses of the phrase) crunching data with various algorithms and models to extract conclusions, insight, and results. And that may be the case to some extent, but there’s also a significant role for AI in localized, focused, sensor-driven applications.

This is demonstrated by the VL53L8CP and similar VL53L7CH 8×8 multizone time-of-flight (ToF) modules from STMicroelectronics. Combined with the company’s new Human Presence Detection (HPD) technology for laptops, PCs, monitors, and accessories, the transducer-based, AI-driven function can accomplish two disparate goals: reduce daily power consumption by 20% in a typical day while offering PC/laptop security and privacy.

Key Features of the New ToF Modules

Along with human presence detection, it integrates features such as multi-person detection and head orientation tracking, as well as advanced functionality like gesture recognition, hand posture recognition, and wellness monitoring through human posture analysis.

Some of the interaction features of this Human Presence Detection scheme may not be immediately obvious or well known:

• Adaptive Screen Dimming tracks head orientation to dim the screen when the user isn’t looking, reducing power consumption by more than 20%.
• Walk-Away Lock & Wake-on-Attention automatically locks the device when the user leaves and wakes up upon return, improving security and convenience.
• Multi-Person Detection alerts the user if someone is looking over their shoulder, thus enhancing privacy.

Let’s Talk Hardware

While all of this is possible due to AI algorithms, it begins with the ToF transducers (emitter and sensor) and data capture. The VL53L8CP ToF 8×8-zone sensor is engineered to optimize the signal-to-noise ratio (SNR) per zone along with its other electrical and optical attributes. At the basic optical level, the device sends 940-nm pulses sourced by a laser diode, with the reflectance “echo” sensed by a single photon avalanche diode (SPAD) detection array (Fig. 1).

These multizone sensors provide an output of raw compact and normalized histogram (CNH) data that AI engines can use for enhanced depth perception, motion detection, and environmental analysis. (CNH data is a measure of the return rate of the photons received by the sensor.¹)

The CNH data format of the VL53L8CH and similar VL53L7CH sensors offers major advantages for data-intensive AI applications. Raw IR signal data from up to 64 (8×8) zones is collected in a maximum of 128 CNH data bins with configurable ranges or widths. All of the CNH data is transmitted to the host via I2C or SPI interfaces, along with the normal ranging distance, signal level, and reflectance data processed by the ToF sensor.

These compact sensors (6.4 × 3.0 × 1.75/1.6 mm) are easy to integrate and support multiple power-supply options. Their small form factor makes them a good fit for embedding behind a variety of cover-glass materials.

The VL53L8CH features a wide 65° field of view (FOV) (45° × 45°); operation up to 400 centimeters (cm) in dark conditions and up to 285 cm ranging under ambient light; and low power consumption (down to 1.65 mW). The somewhat similar VL53L7CH offers an ultra-wide 90° FOV (60° × 60°) and range up to 350 cm in dark conditions, at higher dissipation of 8.3 mW. Both offer multitarget detection and distance measurement in each zone, motion detection with advanced indicators, immunity to cover-glass crosstalk beyond 60 cm, and frame-rate capability up to 30 Hz (Fig. 2).

Various options are available to developers for VL53L8CH and VL53L7CH ToF sensor evaluation (Fig. 3). These include:

• Complete STM32 Nucleo kit with STM32 Nucleo platform and ToF sensor expansion board
• Dedicated ToF expansion board
• Miniature breakout board that can be integrated into customer designs for testing in real environments
• GUI to facilitate evaluation and boost application development

When you step back and realize that a one-centimeter round-trip ToF takes just 67 ps and the resulting system timing precision that figure mandates, this is all pretty amazing. I suspect that many clever engineers will find creative uses for these ToF modules and their associated AI algorithms beyond those offered here, and they will develop innovative applications not yet envisioned or as solutions to long-standing known problems.

References

1. “Frequently asked questions on compact and normalized histograms (CNH)” (TN1527), STMicroelectronics.

2. “Time-of-Flight Sensors,” STMicroelectronics.

3. “Time-of-Flight sensors” PDF documentation, STMicroelectronics.

4. “8×8 multizone Time-of-Flight sensors enabling Artificial Intelligence (AI),” STMicroelectronics.