AbstractAn adaptive algorithm is proposed that can reduce the canceling error at the target point in multichannel active noise control by direct adjustment of adaptive control without explicit identification of each FIR path characteristic when all propagation path matrices are unknown. Instead of reducing the actual error, the proposed method introduces two types of virtual error vectors and adjusts three adaptive filter matrices to make these vectors zero. If the virtual error is converged to zero and the coefficients of the adaptive filters are converged to constant values, the errors at the target point can be made zero. Since the method of generating the virtual errors in the case of a single channel cannot be directly extended to a multichannel case, a generation method for virtual error vectors via special operations is clarified and an algorithm is proposed that adjusts the coefficients of each polynomial of the adaptive filter matrix. If the PE property of the main noise is not satisfied, the corresponding filter coefficients in the proposed method do not converge to the true path characteristics. However, since the canceling error can be made zero, the identification of the true path is not needed unlike the conventional method. In addition, dither signals to assure identification capability are not needed. Finally, through numerical examples, the effectiveness of the proposed method is confirmed. © 2005 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 88(12): 13–24, 2005; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ ecjc.20164