AirComm: Your Smartphone Can “Smell” Fine Particles via Light

Fine-grained air quality monitoring is critical for assessing personal exposure to air pollution in indoor micro-environments; however, dedicated air quality sensors are costly and offer limited spatial coverage. In this work, we present AirComm , a smartphone-based air quality sensing system that requires no additional hardware. AirComm builds on the insight that when a smartphone camera is pointed at an ambient LED, the rolling-shutter readout converts the LED's high-frequency flicker into stripe patterns, while airborne particulates perturb both the temporal pattern of these stripes and their harmonic distribution in the frequency domain. Instead of analyzing image appearance, which is often confounded by exposure settings, scene content, and ambient lighting, AirComm interprets the optical camera communication (OCC) channel and infers PM 2.5 by modeling how particulate scattering affects the OCC link. Reliable inference is challenging because real-world recordings of LED rolling-shutter stripes are often degraded by handheld motion, defocus, saturation, and occlusions. Moreover, LED driving patterns, camera readout pipelines, and imaging viewpoints (e.g., distance and viewing angle) vary substantially. AirComm therefore normalizes the extracted indicators with respect to a carefully selected reference to improve consistency across devices and viewpoints. The resulting features are fused by a physics-guided boost model ( PhyBoost ) for robust PM 2.5 inference. We implemented AirComm on unmodified smartphones and evaluated it through both controlled experiments and real-world deployments under diverse lighting conditions. Evaluated against a co-located Sensirion SPS30 reference sensor which has an accuracy of ±5 μg/m 3 + 5%, AirComm achieves an observed root mean squared error (RMSE) of 6.80 μg/m 3 (MAPE = 5.1%), demonstrating performance consistent with that of consumer-grade PM 2.5 monitors. Furthermore, it accurately captures both transient fluctuations and steady-state PM 2.5 levels under user-induced interference such as handheld motion and viewpoint changes. The AirComm system, along with a real-time demonstration video, is available at https://youtu.be/fu3qD_YM2fM.