Blind source separation-semiparametric statistical approach

The semiparametric statistical model is used to formulate the problem of blind source separation. The method of estimating functions is applied to this problem. It is shown that estimation of the mixing matrix or its learning rule version is given by an estimating function. The statistical e ciencies of these algorithms are studied. The main results are as follows 1) The space of all the estimating functions is derived. 2)The space is decomposed into the orthogonal sum of e ective and redundant ancillary parts. 3) The Fisher e cient (that is, asymptotically best) estimating functions are derived. 4) The stability of learning algorithms is studied. EDICS number: SP 6.1.7 Corresponding Author: Shun-ichi Amari, RIKEN FRP, Wako-shi, Hirosawa 2-1, Saitama 351-01, JAPAN fax: +81-48-462-9881 [email protected] Permission to publish this abstract separately is granted.