Network-Independent Clock Offset Estimation Using Correlated Sensor Data

In many distributed sensing use-cases, sensor nodes need to allocate a time-stamp to sensed data. When combining data from different sensors, it is important to have synchronized timestamps. When using wireless sensor networks (WSNs) without dedicated hardware such as a GPS clock, keeping the sensors synchronised via synchronization protocols will increase communication overhead and the resulting clock accuracy will depend on the latency and jitter of the end-to-end wireless communication paths. In order to become independent of the wireless transmission medium, this paper investigates an alternative approach where the sensed signals with spatial correlation are utilized to achieve clock synchronization. A number of methods are proposed that use maximization of the correlation coefficient to estimate the time shift between two sensing time series. Furthermore, it is shown that a jump detection can be applied as a feature extraction step. The performance evaluation results, obtained using a dedicated testbed, show that the proposed solutions are capable of estimating the offset between two sensor nodes, without imposing additional communication strain and without being affected by the network conditions of the WSN.