A Study Of Spatial Spectrum Estimation Techniques For Non-circular Sources

Many modern telecommunication systems or satellite systems encounter with non-circular sources, where amplitude modulated (AM) or BPSK, ASK and PAM modulated signals are often employed, which signals'amplitudes lie on an in-phase line in the constellation. In other words, these non-circular signals are the real-valued sources, as a result their elliptic covariance do not equal zero. Comparing with the circular signals which elliptic covariances are vanishing, the non-circular sources can employ covariance and elliptic covariance; therefore, the non-circular sources have more information than the circular signals. Improved versions of spectral MUSIC, Root MUSIC and ESPRIT the exploit the non-circularity of sources have been proposed, respectively. These proposed algorithms can double the number of available sensors, thus they can improve the resolution capability and the noise robustness of conventional subspace algorithms and can handle more sources than sensors.In this thesis, the concept of non-circular sources and their property was introduced. In addition to, two ways are pointed out, which can employ the property of non-circuar sources to improve the performance of subspace algorithms. Furthermore, spectral MUSIC, Root MUSIC, ESPRIT and unitary ESPRIT that exploit the non-circular sources have been studied and some mistakes of model have been corrected. Moreover, an improved version of MDL criterion detection of the number of non-circular signals has been proposed. By employing Euler'equation and exploiting the property of non-circular sources, what is more, we proposed the real-valued versions of Root MUSIC and ESPRIT. These real-valued algorithms not only reduce the computational complexity, but also remain the performance of non-circular versions of subspace algorithms.