Blind Source Separaation And Edge Detection And Their Applications In The Study Of Cosmic Reionization Signal

The cosmic reionization signal is an important means and with vital significance toexplore the evolution of the universe.21cm signal is the symbolic signal released duringthe process of the reionization, and it has been regard as the cosmic reionization signal withdeeply researched. Due to the signal is extremely weak, the observation is very difficult.The main reason is the foreground signals are too strongly, which include galaxysynchrotron radiation, galaxy free-free radiation, clusters of galaxies and extragalactic radiodiscrete sources.The paper was wrote from the view of signal processing, used blind source separationfor separations of clusters of galaxies and the cosmic reionization signal, meanwhile theimproved B-spline wavelet function edge detection was applied to detect the edges ofclusters. The contents of this paper is divided into two main parts. Firstly, blind sourceseparation methods and edge detections used to separate and detect the clusters. Secondly,FastICA and polynomial algorithms are applied in the separation of21cm signal.Clusters of galaxies are the important source in the foreground signals and the mainobjects in astrophysics field, how to exclude the interference of other foreground signals isone of the topics in clusters research fields. In this paper, the Joint ApproximativeDiagonalization of Eigenmatrix was used to separate the clusters, supposed clusters as theindependent part from other foregrounds. Edge Detection of clusters has significance in thepositioning and morphology judgment of clusters, improved B-spline wavelet edgedetection algorithm was applied in the edge detection of clusters and the results show thatthe adaptive threshold B-spline wavelet edge detection is better than classic algorithms asRoberts、Sobel、Prewitt.The extraction of21cm signal has always been a problem in the study of cosmicreionization signal. At this stage, in the frequency domain space, the21cm signal extractionis mainly fitted the foreground signals. In the frequency domain space, polynomial fitting,exponential fitting, Fourier fitting, Gaussian fitting, exponent fitting and FastICA were usedto extract the21cm signal. The experimental results show that the performance of theFastICA algorithm is significantly better than the polynomial fitting methods.