A multiple-signal-classification (MUSIC) approach of estimating the code timings of a desired user is considered for direct-sequence code-division multiple-access (DS-CDMA) multipath-fading channels when exploiting multiple receive antennas with either spatially uncorrelated or fully correlated fading. The acquisition performance of the conventional MUSIC timing estimator employing a single antenna is not good for the small size of observation windows and low signal-to-noise ratios (SNRs). Multiple antennas allow for rapid acquisition and lowers the range of detectable SNR. An efficient and improved MUSIC algorithm of estimating the multipath timings of a desired user for DS-CDMA systems is presented. In multipath-fading channels, the solution of the proposed algorithm is based on successively optimizing the cost function for increasing numbers of multipath delays, which does not require a multidimensional search for multidelay paths. Furthermore, the estimate of code timing at each path is obtained by finding the zeros of second-order polynomials, which is computationally efficient. The proposed MUSIC algorithm significantly improves the acquisition performance of conventional MUSIC algorithm in the presence of multipath time-varying Rayleigh-fading channels with arbitrary time delays. The acquisition performance of multiple antennas-based MUSIC timing estimators is much better than that of a single-antenna-based timing estimator. The Crame´r-Rao bound for the code-timing estimator based on multiple antennas is presented.