Abstract: In this paper, a new observation function of BLAST-OFDM system under non-gaussian noise is proposed, which is used for frequency selective fading channels and non-gaussian noise in application environment of BLAST system. With QR decomposition of the channel matrix, the static observation function of BLAST-OFDM system is transformed into a dynamic state space model, and then the particle filter is used for space-frequency layered detection. Making the full use of the finite alphabet of the digital modulated communication signal, the optimal proposal distribution can be chosen to produce particle and update the weight. When incorporated with current method of reducing error propagation, a new space-time layered detection algorithm is proposed. Simulation result shows the validity of the proposed algorithm.