The invention relates to a particle filtering method used for asynchronous DS-CDMA system blind multiuser detection in the mobile communication technical field to inhibit multiple access interference and near-far effect and reduce system error rate. The method adopts the following main steps: firstly, a plurality of channel gains, time delays and random particles of a user symbol are generated, and the importance weighing of the random particles is initialized; secondly, a Markov Model, which is set up according to the characteristics of the physical channel of each user, combines with user symbol particle to complete Kalman filtering estimation of each channel gain and time-delayed particle, thereby reducing the deviation of particle distribution; thirdly, the joint posterior probabilityof the particle is obtained and the weighted mean of the particle in the estimated value of the user symbol, channel time delay and gain is completed; fourthly, during the process, update and normalization of particle weighting are completed, and importance resampling of the particle is carried out according to effective sampling size. The invention has the characteristics of stable performance and easy implementation.