Abstract: For the purpose of recognizing unitary space-time signals in fading channels, we propose two identification schemes. One is maximum likelihood recognition scheme which employs channel transfer probability density function (PDF) to construct average likelihood ratio and generalized likelihood ratio classification functions. This scheme finishes classification with the difference of the likelihood ratios of different space-time signals, and its calculation complexity can be decreased by processing in log domain. Another is high order statistics recognition scheme which uses moment generating function to generate high order joint moment and high order joint cumulant. This scheme realizes recognition by the special high order statistics of unitary space-time signals. The proposed schemes can identify unitary space-time signal from conventional space-time codes. Maximum likelihood recognition scheme can realize identification without channel state information (CSI) and can give a better performance on a large scale when CSI is available. The scheme based on high order statistics needs CSI and its performance is affected by the accuracy of channel estimation. Compared with the maximum likelihood method, high order statistics scheme shows a relative worse performance but its calculation complexity is low. In both schemes, the recognition performance can be improved by increasing the number of receive antennas. Compared with 2 receive antennas, the gain for 4 receive antennas is 7-10dB for maximum likelihood method without CSI and is about 45dB for high order statistics method with CSI. Simulation shows the effectiveness of the proposed schemes.