Automatic calibration of sensor-phones using gaussian processes

In the context of environmental sensors embedded into location-aware cell phones, this paper explores how to automatically calibrate the bias of each sensor based on readings from nearby phones, thus exploiting the mobility of the sensors. Gaussian process regression is used interpolate values and also infer the bias of the sensors. The impact of dense vs sparse sampling is explored, and future strategies for efficiency, gain calibration, and cyclical covariance are suggested.