Multiple Source Data Association for Distributed Acoustic Sensor Network in Open Environment
Muhammad Saad Ayub, Jianfeng Chen, Anam Zaman · 2021
In this paper, we investigate the problem of data association occurring in the localization of multiple sound sources for a distributed acoustic network placed in an outdoor environment. Multiple source signals satisfy the assumption of window-disjoint-orthogonality (WDO) and the direction of arrival (DOA) of each signal is extracted. The data association problem is imminent when we receive multiple DOAs from different microphone arrays in the network and our central node is ambiguous about the DOAs that belong to the same source. A methodology is presented for accurate association and location estimation of multiple sources based on the Bayesian network. In a realistic environment with reverberation and noise usually, the WDO condition of signals is not satisfied and the system misses the detection of some of the signals. To overcome this issue the information of distance between the arrays is incorporated in the network such that the arrays closer to each other are more likely to detect the same acoustic signal. The methodology is tested for non-speech signals using simulations and outdoor data experiments. The results confirm that the methodology achieves higher association and localization accuracy as compared to existing methods.