A new kind of blind separation algorithm for real-time linear mixture signals is presented in this paper.At first,the mixture matrix is changed into orthogonal matrix by using whitened mixture signals.Then based on QR decomposing theory,a series of primary rotating transforms are performed for mixture signals.It is also considered that the mutual information is the least if source signals are independent and a new kind of adaptive blind separation algorithm is educed.This method avoids the complex calculation of |det w| with the present information theory method (e.g.Torkkola 1996,Pham 1999,Lee 2000,as well as Tang 2000).In the paper we not only give the theoretic proof in detail,but also carry out the simulation experiments.Both of theoretic analysis and simulation results show that this algorithm saves the separation time and has a good separation effect.