分布式平均共识的无线声传感器网络时钟校准

Clock Synchronization for Wireless Acoustic Sensor Networks via Distributed Average Consensus

在无线声传感器网络中,每个节点的时钟系统通常独立,受制造工艺、温度、湿度等因素的影响,各节点间的时钟存在异步问题。时钟校准方法是补偿节点间的采样率偏差的有效手段。然而,大部分已有方法只能以集中式的方式适用于无线声传感器网络中,这不仅需要额外的中心处理单元,也将带来较重的通信负担。针对上述问题,本文提出一种无线声传感器网络时钟校准方法。首先,在线性相位漂移模型中引入解卷绕策略,以消除高频处的相位模糊问题,进而提高每对相邻节点间的时钟同步精度。接着,基于分布式平均共识算法,提出一种分布式时钟校准新方法,使各节点并行地、协作地进行时钟同步。其次,在补偿各节点采样率偏差的基础上,仅利用到达时间信息进行分布式声源定位。最后,根据声源位置,分布式地估计各节点之间的起始采样偏差。与已有方法不同,所提方法仅需相邻节点之间的局部通信、无须中心处理单元、鲁棒性强,更适用于无线声传感器网络。本文实验结果表明了该方法的有效性。

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.