Efficient And Realtime DOA Estimation Of Low-cost Wireless Acoustic Sensor Array Network

Wireless acoustic sensor array networks(WASANs)are the combination of wireless acoustic sensor network(WASN)and sensor array processing.Therefore,it has all their advantages such as high accuracy,self-organization and strong concealment,and has been widely applied for the passive targets localization of low latitude UAV,ground heavy vehicle and underwater submarine in the military area.However,several resource limitations,such as local computational capac-ity,communication bandwidth,power supply and even cost,make efficient and realtime DOA estimation on low-cost wireless acoustic sensor array a big challenging.The newly emerged com-pressive sensing(CS)technology can reconstruct the raw signal with overwhelming probability from reduced random measurement of the raw signal when the signal can be sparse represented in a certain domain.It has great potential for WASAN system,and solves the multiple resource restriction of WAS AN platform.Based on the overview of some domestic and overseas research of high efficient,real-time source localization.This thesis focuses on the multiple resource restriction of signal sampling,data transmission,signal processing in WASAN system and solves them by introducing the CS technology to reduce resource requirements including signal sampling energy,array processing redundancy,data transmission rate,and proposes a fast DOA estimation algorith-m for the low power,low computational capacity sensor nodes.The main content of this thesis is summarized as follows:1.A random subsampling approach is proposed for array data acquisition.Based on the frequency domain sparsity of source signal and the resource limitations of WASAN system,a random subsampling approach is designed to take measurements with reduced dimension,which greatly reduces the array date volume without special purpose hardware designing or complicate local processing task.2.A joint sparse representation based DOA estimation approach is proposed.It utilizes the joint sparse property of array data in both frequency domain and angle domain,and uses a joint sparse representation approach to directly estimate the DOA spectrum from the random subsam-pled array data without reconstructing the raw array signals.Therefore,the proposed approach reduces the array data volume and avoids the error propagation from reconstructed signal to DOA estimation in conventional compressive sampling-signal recovery-DoA estimation framework.3.A one-bit quantization based DOA estimation approach is proposed.Based on the obser-vation that DOA estimation focuses on the time delay among sensor nodes,the proposed one-bit quantization based DOA estimation approach exploits the waveform reservation of one-bit com-pressive sensing technology and fulfills direct DOA spectrum reconstruction with only sign bit measurements in local sensor nodes.4.A fast direct 3D direction of arrival(DOA)approach is proposed.It uses array geometry to solve the phase unwrapping issue of widely space sensors and calculates the 3D DOA result in the phase space by using a linear 3D DOA estimation model.Therefore,the large computational complexity issue for some brute force search technologies such as AML,MUSIC approaches can be solved.Theoretic analysis proves that the error covariance of the proposed approach is the same as the corresponding Cramer-Rao lower bound(CRLB).5.A fast phase measurement based distributed source localization approach is proposed.The proposed approach only needs to transmit some pre-processing results to the remote fusion center without large data volume transmission.Also,the proposed approach has the same error covariance as the CRLB of the conventional complete information based source localization approach.