DSP Based Implementation of Direction of Arrival for Wideband Sources

Abstract

There is a need for the computation of direction of arrival (DOA) for wideband sources for number of applications. There are number of algorithms available in the literature for wideband case however we focus on Coherent Signal Subspace method proposed by Wang and Kaveh. Most algorithms available in the literature follow some variation of CSS algorithm thereby increasing computational complexity in an effort to get more accurate, statistically stable and unbiased DOA. We are interested in computing DOA for wideband sources in real time. We chose CSS algorithm and investigated possibility of implementing it using commercially available digital signal processor (DSP) in an effort to achieve real time capability. DSPs offer flexibility, ease of development of embedded system, reduces design cost and offers use of high-level programming language such as C. In this work, we propose a DSP based architecture for detecting and estimating the DOA of wideband sources. It is known that DOA algorithms require computation of eigenvalues and eigenvectors. It would be best to find computational friendly algorithm for the computation of eigenvalues and eigenvectors. In this work eigenvalues and eigenvectors are computed using well known Householder and QR algorithms. CSS algorithm is then implemented in C and executed on DIOPSIStrade 740 by Atmel. DIOPSIStrade 740 (D740) is a high performance dual-core processing platform for real time applications. The CSS algorithm was then parallelized and a parallel architecture was then developed. This paper presents parallel architecture using DIOPSIStrade 740 (D740) and computes performance parameters.