Antenna switch enables multiple antennas to share a common RF chain, thus an additional spatial dimension, i.e., antenna index, can be utilized in the design of single RF chain MIMO and information can be conveyed via both signal space and spatial dimension. In this paper, we propose a unified adaptive transmission scheme - adaptive spatial modulation that allocates information into signal space and spatial dimension in order to maximize the overall channel capacity for single RF chain MIMO. The proposed adaptive spatial modulation is realized by using Huffman coding, i.e., designing variable length codes to activate the transmit antenna with different probabilities. The optimal antenna activation probability is derived through optimizing channel capacity. To make the optimization tractable, closed form upper bound and lower bound are derived as the effective approximations for channel capacity. Numerical results show that the proposed adaptive spatial modulation offers considerable performance improvement over both spatial modulation and transmit antenna selection.