The Application Of MP In The Single Channel Blind Signal Separation

Sometimes, due to system complexity and cost, the number of sensors which used for receiving signals are limited. Especially using only one sensor to recover the sources signal which is called Single-Channel Blind Signal Separation(SCBSS) is one of the hot research directions. Because the problem of SCBSS that only having one sensor is very special, many original separation methods are restricted. To solve this problem, we use the Sparse Decomposition based on Matching Pursuit(MP) to do our research. In this paper, we proved that the MP can be used in SCBSS through theoretical analysis. And then applying MP sparse decomposition to single-channel acoustic logging signal separation in order to prove the actual value of MP. The following four parts are mainly included in this paper:1,The existing methods of single channel separation are classified and summarized, On this basis,MP is used to the single channel blind signal separation2,Because sparse decomposition is based on the over-complete dictionary, firstly over-complete dictionary is created through dispersing the parameter of generating function. And then we use probe to probe actual original signal in different libraries to find the best matched atom and construct the over-complete dictionary by using MP sparse decomposition.3,Firstly, according to the characteristics of the actual logging signal, single-channel artificial logging signals are compounded. Secondly, single-channel artificial logging signal are separated by MP sparse decomposition,and the result is effective, On this basis, different noise is added to the artificial logging signal to measure the anti-noise performance of the algorithm by calculating the reconstruction error. Through computer simulation, we confirming that it is feasible of using MP sparse decomposition to the single-channel blind signal separation. After that, the MP sparse decomposition is applied in practical single-channel logging signal to see whether the algorithm have the same efficiency to the practical signal.