QIF-T0102

Passive facial geometry estimation via display-as-illuminator inverse photometry (presence/orientation detection without camera)

The device display functions as a structured light source — each frame emits a known photon distribution. Reflected light from the user's face is captured by the ambient light sensor (ALS). The known spectral emission S(x,y,λ) and measured reflected irradiance E_sensor = ∫∫ ρ(x,y)cos(θ)S(x,y,λ)/r² dA constrain an inverse photometry problem. CRITICAL FEASIBILITY CAVEAT: A single ALS integrates the entire reflected light field into one scalar value, making 3D geometric reconstruction an ill-posed inverse problem. With current single-sensor hardware, achievable resolution is limited to basic presence detection, head orientation, and coarse proximity estimation — NOT high-fidelity facial geometry. Identity matching might be feasible only against a small template library with strong a priori constraints. Future multi-pixel ALS or multi-sensor arrays could significantly improve reconstruction fidelity. Despite geometric limitations, the same ALS reliably detects physiological signals: pulse-modulated skin reflectance for PPG heart rate (demonstrated in Samsung Galaxy phones) and skin color variations for SpO2 estimation.