Sony Patent Keeps Face Tracking Active After the Camera Loses Sight
If you lean out of frame or a shadow falls across your face, most camera-based face tracking just gives up. Sony's new patent describes a system that hands off to a motion sensor you're wearing, so tracking never drops.
What Sony's camera-plus-motion-sensor tracking actually does
You're watching a big screen TV and the system is adjusting the picture based on where you're sitting and which way you're facing. Then you lean sideways to grab something, or the room light shifts, and the camera can't find your face anymore. The tracking stops, and the experience breaks.
Sony's patent describes a fallback: a small motion sensor worn by the viewer (think a smartwatch or fitness tracker) that can step in the moment the camera loses your face. The camera stays in charge when it can see you. But the moment face recognition fails, the system switches to the motion sensor's data to keep estimating where your head is pointing.
The two sources work together, with the camera taking priority and the wearable acting as a safety net. The result is tracking that stays continuous even when the camera's view is blocked or too dark.
… continues the estimation of the position and posture of the face of the user on a basis of the IMU data in a case where it is determined that the execution of the face recognition processing has failed.
Translation: When the camera loses sight of your face, the headset sensors take over to keep tracking you.
How the system switches from camera to wrist sensor mid-session
The system combines two separate inputs: a camera fixed to a display (like a TV or monitor) and an IMU (inertial measurement unit) worn by the person watching. An IMU is essentially the same kind of motion chip inside a phone or fitness band, measuring acceleration and rotation to figure out which direction something is pointing.
Normally, the camera runs face recognition processing to estimate the viewer's head position and orientation in real time. This works well when the camera has a clear, well-lit view of the viewer's face.
When that face recognition fails (because the viewer moved out of frame, turned away, or lighting conditions dropped), the system doesn't wait for the camera to recover. Instead, it automatically continues estimating head position using the IMU data alone.
Key behaviors the patent describes:
- Camera-based face tracking runs as the primary method
- The system actively checks whether each tracking attempt succeeded or failed
- On failure, IMU data from the wearable takes over immediately
- Camera tracking resumes priority once it succeeds again
The patent doesn't specify how the two streams are fused when both are available, but the core logic is a clear primary-plus-fallback architecture.
What this means for TV viewing and display-based eye tracking
Display systems that track where a viewer is looking or sitting are used for things like auto-adjusting 3D depth on a glasses-free 3D TV, optimizing sound positioning, or personalizing a picture based on viewing angle. All of those break the moment tracking drops, and camera-only tracking drops often in real living rooms where people move around, dim the lights, or sit off to the side.
A wearable fallback is a practical fix to a real gap. Sony's run of display-and-viewer-tracking filings suggests the company is working toward displays that feel more aware of who's in the room. Whether that means a next-generation Bravia TV or something in the PlayStation ecosystem, continuous head tracking without requiring the viewer to stay perfectly still is a meaningful step.
That makes this Sony's 19th filing we've tracked since July in the AR glasses race, a run that includes applications like gaze-directed remote cameras and floating image sunglasses.
Camera-based face tracking fails constantly in the conditions where people actually watch TV: dim lighting, off-center seating, a viewer who keeps moving. Every time tracking drops, any feature built on top of it breaks too, which is why depth-adjusting displays and gaze-based menus remain curiosities rather than shipping products.
Sony's answer is to use the motion sensor already sitting on a viewer's wrist as a backup when the camera loses the face. The engineering challenge is the handoff logic between the two, and that turns out to be a tractable problem rather than a research-level one.
The approach matches the size of the problem only if the viewer is already wearing something. That assumption holds well in gaming or fitness contexts, but is harder to guarantee for a household TV, and how Sony closes that gap will determine how much of this patent's value actually reaches consumers.
There are more where this came from
We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.
The drawings
9 drawing sheets from US 2026/0261642 A1 · click any drawing to enlarge
Want this weekly breakdown for a company we don't cover? Patentlyze Pro →