| In recent years, blind signal processing (BSP) is one of the hot topics in the study ofsignal processing. The number of source is unknown, that limits the exactness of blind signalprocessing vastly, so it is an urgent problem need to be solved in blind signal processing. Thispaper mainly aims at the case of owe multi-channel and single channel to research theproblem of blind source number estimation.Firstly, in this paper, we describe the research background and the homologousmathematical model of blind source processing, the domestic and overseas research present ofblind source number estimation. Through the experiment and the theorrtical analysis, wediscuss the source number estimation influence on MUSIC algorithm, analyze advantages anddisadvantages of the main algorithms of existent source number estimation, propose MUSICalgorithm base on eigenvalue decompose under the case of unknow source number, and thebetter property of this algorithm is verified by experiment.Secondly, we introduce the problem of owe multi-channel blind signal, analyze theapplication of signal sparse processing technic in source number estimation, propose signalsparsity source number estimation algorithm base on K′mean value, realize the accurateestimation of blind mixed signal source number under the condition of owe multi-channel.Finally, we analyze the particularity of single channel blind signal, introduce a problemof single channel blind signal dimension extend base on filter. Aim at single channel blindsignal with noise, combine with four-order cumulant algorithm and self-adapt thresholdeigenvalue decompose algorithm, propose a number estimation algorithm of single channelsource with noise base on high order cumulant, and estimate the number of single channelsource with noise effectivelly. |
