Aiming at the problem of high complexity of particle-filtering in single channel blind separation for digital mixed signals, a novel low complexity algorithm is proposed. By analyzing, the high algorithm complexity lies in the exponentially increase between the number of researching spaces and the smoothing length in particle sampling process. Two sampling schemes are proposed, which are named partial sampling and hybrid sampling. The basic ideal of partial sampling is that instead of exploring the whole space of smoothing interval, several symbol sequences are generated. The incremental weights of each symbol sequences are used for particle sampling and weight update. The hybrid sampling scheme divides the smoothing interval into two parts, in which the first part uses the traditional full space searching scheme and the second one uses the partial sampling scheme. The incremental weights of the two parts are used for particle sampling and weight update. Theoretical analysis and simulation results show that the proposed algorithm can efficiently reduce the particles’ searching spaces.