For Orthogonal Frequency Division Multiplexing (OFDM) multiple-input multiple-output (MIMO) multiplexing in high speed wireless transmission under broadband fading channel, a novel coded layered space-time-frequency (LSTF) architecture is proposed, which can achieve the available spatial, temporal and frequency diversity, and make the system implementation easy for high data rate transmission. In this novel architecture, each independently coded layer is threaded in the three-dimensional space-time-frequency transmission resource array. Channel estimation based on a time-multiplexed pilot channel is employed. The convolutional code is used as the constituent code, and the iterative receiver structure consisting of a linear minimum-mean-square-error soft-interference- cancellation (LMMSE-SIC) detector and the soft-input soft-output (SISO) maximum a posteriori (MAP) decoders is adopted. Simulation results show that the proposed LSTF architecture can considerably outperform the LSTF (i.e., LSTF-b) in which each independently coded layer is associated with a permanently assigned antenna, and get almost the same performance as the LSTF (i.e., LSTF-a) where coding is applied across the whole information stream. Since its structure consists of multiple parallel lower-speed encoders/decoders with shorter codeword length, the proposed LSTF architecture can be much more easily implemented than the LSTF-a. Moreover, four iterations for joint detection and decoders are sufficient, and the percentage of the pilot channel energy in one packet should be in the range of [16%, 24%] to provide near optimum performance.