Acoustic Sensor Array System For Real-time Estimation Of Direction-of-Arrival

Using acoustic array to estimate the direction of arrival for targets has been one of the hot spots in the research and application. In particular, due to the advantages the acoustic sensor arrays have, such as easy to hide, a good anti-interference ability and small power consumption, make them applicable in some special area where it’s inconvenient to enter for human, such as battlefield and forest etc. However, with the conception of digital battlefield and intellectualized security, acoustic array is facing new challenges. It is a new problem that how to design and complete a high accurate, low power consumption, real time and be able to work independently and stably for a long time.Considering such a problem, this article uses an existed non-real-time system to discuss the performance of AML. Then the article offers approaches for the real-time system based on the algorithm and system. They are decreasing the DOA estimation algorithm, high expensive real-time system and running the DOA estimation algorithm on the PC. This article includes four main parts:1) Using a real system to prove and discuss the AML (Approximated Maximum Likelihood) algorithm, the simulations prove that the AML algorithm has ultra-high resolution and small estimated deviation; Utilize the non-real-time Voxnet system to check the validity of the AML algorithm in a real system, a series of tests prove AML algorithm still be able to maintain a high estimated performance in the low SNR environment; Furthermore, by using the DOA results got from different nodes, locate the target and the error of the localization is less than30cm with4nodes. In addition, a source separation application based on the AML algorithm is also present in this chapter. Finally, introduce the ways to implement the real-time acoustic sensor array system.2) For decreasing the complexity of DOA estimation algorithm purpose, propose a new low complexity DOA estimation method based on regularized orthogonal matching pursuit (TROMPDOA). It is shown by simulations that compared to the previous orthogonal matching pursuit DOA (OMPDOA) estimation method, the new method has better stability, especially under the low SNR and closely spaced signals conditions. In addition, the new method has a similar or even lower complexity than OMPDOA method. At last, by running a test based on the real acoustic signal, prove the new method has a similar performance to Capon beamforming method, while better than OMPDOA method in a real system platform.3) Design and implement a real-time acoustic sensor array node system based on the ARM+FPGA+DSP architecture. The system is used to collect data, local algorithm process and transfer the DOA result to the fusion center in an acoustic sensor array system including19sensors. In the outdoor experiment, the system estimates the DOA of targets and locate and track the targets by fusing the DOA results from different nodes.4) Finally, according to the disadvantages the traditional acoustic sensor arrays have, such as difficulty of arrangement and high power consumption etc, propose the conception of virtual wireless acoustic sensor array system. By using the compressive sensing and transfer the original data to confusion center to run the DOA algorithm.