Adaptive Beamforming Of Interference Covariance Matrix Reconstruction Based On K-means Clustering

Beamforming can be used for sound source localization through microphone sensor arrays. Adaptive beamforming is extremely sensitive to model mismatch, particularly when the desired signal is present in the training data, which has a significant impact on beamforming efficiency. In order to address this problem, an adaptive beamforming algorithm based on K-means clustering and INCM (interference plus noise covariance matrix) reconstruction is proposed in this paper. When the expected signal and the interfering signal have a large difference in power, the expected signal and the interfering signal are identified using the K-means clustering algorithm, and the expected signal components are extracted from the covariance matrix to reconstruct the INCM. Meanwhile, a QCQP (quadratic constrained quadratic programming model) is constructed to improve the output performance degradation of the beamformer when the steering vectors of the expected signal are not matched. The results show that the proposed algorithm suppresses interference output more effectively while retaining a high output SINR (signal to interference plus noise ratio).