Abstract: Spatial modulation (SM) has received much attention recently as a novel multi-antenna technology. The spatial modulation with single-active antenna has been expanded to generalized spatial modulation with multi-active antenna, further improving the transmission rate. However, this makes the demodulation more difficult. The complexity of ML detection algorithm will increase exponentially with the augment of activated antenna, and then the ML algorithm becomes a NP-hard problem. To solve these issues, we exploited the inherent sparse property of the spatial modulation signal and proposed a sub-optimal detection algorithm. Our numerical experiments show that the proposed algorithm can guarantee the detection performance as well as reducing the complexity. Moreover, due to the use of sparsity, our proposed approach can also be applied in the underdetermined system.