When the MUSIC (multiple signal classification) algorithm is used by the orthogonal frequency division multiplexing (OFDM) system to perform time delay estimation under the condition of low SNR (signal-noise ratio),the signal subspace and the noise subspace obtained by eigenvalue decomposition may not be completely orthogonal,which leads to the MUSIC algorithm performance degradation.In order to solve the problem,a time delay estimation algorithm based on the weighted subspace—WMUSIC (weighted multiple signal classification) algorithm was proposed.The proposed algorithm performed eigen-decomposition of the covariance matrix of channel frequency domain response estimation,weighted the noise subspace by the power series of the noise eigenvalues,and at the same time used the reciprocals of the signal eigenvalues to weight the signal subspace to correct the pseudo-spectrum of the MUSIC algorithm and obtained the delay estimation value through spectral peak search.The simulation results show that,compared with MUSIC algorithm,WMUSIC algorithm has higher time delay estimation accuracy.Specifically,when the SNR is －8 dB,the estimation accuracy can be improved by 71.46% with two paths and 19.48% with three paths.The WMUSIC algorithm effectively solves the problem of relatively large errors in spectral peak search caused by spectral peaks aliasing for MUSIC algorithm with low SNR.The proposed WMUSIC can complete more accurate time delay estimation,which provides a reference for improving the performance of time delay estimation in OFDM system under low SNR conditions.