| Summary: | 博士 === 國立成功大學 === 電機工程學系碩博士班 === 98 === The problem of the direction-of-arrival (DOA) is to estimate the directions of multiple incident signals from noisy measurements received at a sensor array. For practical DOA estimation applications, the conventional subspace-based method has some inherent problems. Recently, Kim and Wen proposed pseudocovariance matrix techniques to improve the performance of subspace-based algorithms in the practical DOA estimation applications. But they still are against real-time requirement due to high computation complexity.
In this dissertation, we proposed two different approaches with application to the uniform linear sensor array (ULA) in a single snapshot case to satisfy the demands of practical applications for conventional subspace algorithms. The first part of our approaches presents a novel fast DOA algorithm, using an orthogonal projection and noise pseudo-eigenvector technique with forward-backward data model. The second part of our approaches presents another novel fast DOA algorithm, using the technologies of the shrinking signal subspace and the noise pseudo-eigenvector with forward data model. Without eigendecomposition computation and employing shrinking signal subspace technique, our proposed approaches can reduce computational complexity(the computational complexity of Kim’s and Wen’s methods is while ours is , where denotes the number of sensors) while maintaining better or similar resolution capability when contrasted to Kim’s and Wen’s fast DOA estimation algorithms.
Simulation results showed that for SNR=5 dB, three uncorrelated signals from [2.3,9.7,17.8] and =23, the resolution probability of the first approach and Wen’s method were about 0.8 while Kim’s method is less than 0.6, and for SNR=5 dB, two coherent signals from [5,10.5] and one uncorrelated signal from [16] and =22, the resolution probability of the second approach was close to 0.2 while Kim’s method was close to 0.
|
|---|
