Clock Synchronization for Wireless Acoustic Sensor Networks via Distributed Average Consensus

‍ ‍In wireless acoustic sensor networks， the clock system of each node was independent， resulting in the asynchronous clocks among nodes due to the manufacturing tolerance， temperature， humidity， and etc. To this end， some time synchronization methods have been proposed recently. However， most of the existing methods could only be adopted for wireless acoustic sensor networks in a centralized manner， which requires an additional central processor and a large amount of data communication. To mitigate the above problem， a distributed clock synchronization algorithm was proposed in this paper. Specifically， the unwrapping strategy was first introduced in the linear-phase drift model to eliminate the phase ambiguity that appears in high-frequency bins， from which the different sampling rate offset （SRO） between each pair of nodes was recovered. Then， a distributed SRO estimation method was derived by introducing the distributed average consensus algorithm， where all nodes work in parallel and cooperatively. Next， the sound source was localized by using only direction-of-arrival（DoA） measurements. Finally， according to the estimated source position， another distributed approach was presented to obtain the initial sampling bias among different sensor nodes. Unlike existing methods， the proposed method requires only local communication among neighboring nodes， and does not require a central processing unit， thus it is more suitable for wireless acoustic sensor networks. Experimental results confirm the validity of the proposed method.