Abstract: The performance of the traditional generalized sidelobe canceller algorithm decreases significantly in complex noise environments. To improve its applicability, we propose an improved generalized sidelobe canceller algorithm based on linear microphone arrays. This method replaces the fixed beamformer of the generalized sidelobe canceller algorithm with a minimum variance distortion-free response beamformer, and enhances the target speech signal more effectively and suppresses noise interference in other directions. An adaptive filtering algorithm with an improved blocking matrix is used to obtain more accurate estimate of the reference noise. Finally, residual noise is further eliminated using an optimally modified logarithmic spectral amplitude algorithm. Experiments by recording speech signals in various noise environments are conducted. The results demonstrate that the proposed algorithm achieves superior noise reduction performance and higher speech quality compared to other algorithms.