Research And Implementation Of MUSIC Algorithm Based On Multi-Core DSP

Direction of Arrival is one of the important contents of array signal processing.Multiple Signal Classification(MUSIC)algorithm is an important method of DOA estimation,which has the advantages of super resolution and high precision,and has great value in engineering implementation.However,because of the complexity of music algorithm,the long time in hardware implementation and the strong dependence on the array structure,it is a great challenge for hardware implementation.MUSIC consists of two primary modules,eigenvalue decomposition and spatial spectrum search,generally consume a significant hardware and cause long processing delay,so the improvement of music algorithm is mainly focused on these two aspects.The fast decomposition algorithm,such as gradient algorithm and power iterative algorithm,is proposed for eigenvalue decomposition.The fast search algorithm,such as dimensionality reduction search algorithm,symmetric compression search algorithm and search-free algorithm,is proposed for spatial spectrum peak search.In this paper,we design and develop an efficient parallel implementation of DOA on DSP TMS320C6678 for 2-D DOA estimation.As mentioned,MUSIC consists of two steps: covariance decomposition and peak searching of the spatial spectrum.Covariance decomposition: we use Householder and Givens to transform the original complex domain into the real domain.Peak searching: the peak search uses the hierarchical search,symmetric compression search,and allocates it to multiple cores.For hardware,we choose TI's powerful TMS320C6678 multi-core DSP chip,using the BIOS,the inter processor communication and shared memory modules,to achieve a multi-core parallel computing.In the case of high real-time requirements,this method has important value.