A class of multichannel sparse linear prediction algorithms for time delay estimation of speech sources

•It develops a multichannel sparse linear prediction approach to time delay estimation of acoustics sources in reverberant and noisy environments, which can be viewed as a generalization of the previously developed multichannel cross-correlation coefficient (MCCC) algorithm and the multichannel spatio-temporal prediction (MCSTP) algorithm.•Based on the sparsity of the prediction coefficient matrix of speech signals, a class of algorithms are developed for TDE, including the multichannel spatio-temporal sparse prediction and the multichannel spatio-temporal group sparse prediction algorithms.•Through theoretical analysis and simulations, it is shown that the developed TDE algorithms are more robust to noise and reverberation than the MCCC and MCSTP algorithms.