Co-channel signals tracking is an important topic in the field of radio monitoring. A direct tracking method based on random spatial spectrum is proposed for multiple mobile co-channel transmitters. The co-channel signals completely overlap in both time and frequency domains, but can be separated in the spatial domain. The random spatial spectrum is derived based on a set of state particles. The spatial sparse method is then proposed to track the positions based on the sparsity of a grid spatial spectrum. The algorithm is verified by the constant turn rate and acceleration motion model. Simulation results are given to demonstrate that the proposed algorithm outperforms other tracking algorithms and provide robust locating performance for multiple co-channel targets with different received powers.