Image-sourced fingerprinting for LED-based indoor tracking

The advent of location-based services has created a necessity for indoor positioning and tracking. We propose an automated fingerprinting-based algorithm for an indoor tracking system that uses light-emitting diodes (LED) equipped with imaging sensors. The fingerprint collection part requires a calibration step that consists of the following: (i) a theoretic light power distribution is calculated using the visible light communication channel model; (ii) the imaging sensors mounted on the LED arrays capture the grayscale images; (iii) the captured images and the theoretically calculated light power distribution are used to estimate the light power map. Linear and non-linear regression, and artificial neural networks are used for mapping from the grayscale pixel values to the light power levels. A photodetector on the user equipment measures the light power in real-time. The estimated fingerprint map and the real-time power measurements are used in an extended Kalman filter. The results show that we can reach an accuracy on the order of the degree of granularity of the fingerprint map.