Abstract: Friedland's separated-bias estimation algorithm for linear systems is extended to a class of nonlinear systems with output bias. Based on the orthogonality principle proposed by the authors, a pseudo separated-bias estimation algorithm for time-variant bias changing in large region is further proposed. These not only develop friedland's separated-bias estimation theory, but also make it easier to apply for the extended Kalman filter. The effectiveness of the proposed algorithm is demonstrated by computer simulation.