MUSIC Algorithm And Its Parallel Implementation

Now, in the field of array signal processing, DOA(Direction of Arrival, DOA)estimation has became to one of the very important issue, and it have a wide range of applications in many fields like radar, wireless communications, sonar, etc. DOA estimation is a technique that be used to determine the direction of the radiation source,and it received information through an array, and then accord to a certain pre-processing methods to estimate the orientation of sources. DOA estimation can be realized by a lot of algorithms, wherein digital beam-forming is represented by the traditional method has a significant drawback is that constrained by the Rayleigh limit.In a word,it has a low estimate accuracy, and can not distinguish a few letters source in a beam width.And later the multiple signal classification(MUSIC) algorithm that is an important representative of the super-resolution algorithm to solve this problem, this algorithms are no longer limited by the Rayleigh limit.This thesis from one of this classic super-resolution algorithms:MUSIC algorithm to analyzes it on the application of DOA estimation.Then we based on that does not affect the estimation of source azimuth to use parallel processing technology to improve their operational computing speed. In this way,we can let the algorithm can be obtained in the field that have an important requirement for real-time.This thesis first introduces the basic theory of MUSIC algorithm,and then analyzed in detail through experiments that the algorithm’s capacity in the issues of DOA estimated processing. And analyzed the influence of MUSIC algorithm in DOA estimation by noise ratio, the array element spacing, etc.In this thesis, covariance matrix of the array be gotten by that correlation process the received signal is a Hermite matrix, then we propose a preprocessing method to this matrix complex to let the Hermite matrix(complex matrix) converted into the corresponding real matrix, which will greatly save processing time and let the algorithm more suitable for implementing in hardware DSP. Also during the same time conducted a forward and backward smoothing process, so this method is not only estimate effectively for the DOA estimation of non-coherent signal source(the mainconsideration of this thesis)but also the coherent sources.Be considered from many aspects of the time-consuming,hardware applicability and parallelism,this paper aim the core of the MUSIC algorithm that is the feature space decomposition to select the QL algorithm in many common features decomposition algorithm.Then detailed analysis serial implementation between each module of the MUSIC algorithm.Finally, the parallel approach that been proposed by this thesis is feasible proved by experiments.And it verify by experimental data that the time-consuming has been signigicantly reduced without affecting the algorithm performance in direction estimation.