| Blind signal identification, as intermediate step between signal detection and decoding, plays a crucial role in non-cooperation MIMO communication scenarios and has been used in many military and civilian field such as spectrum monitoring and surveillance, software defined radios and wireless communication countermeasure. As a crucial problem for blind signal identification, accurate estimation of the number of transmit antennas of MIMO system provides prior knowledge for efficient estimation of the signal type,the channel and the space-time coding mode, and gains great interest of scholars for its study value. Based on the previous works of scholars, the following tasks are made in this thesis.(1) Some basic knowledge of RMT and MIMO system is briefly introduced; the AIC and MDL algorithm based on information theoretic criteria,the PET algorithm based on classical multivariate statistical theory and the IWME algorithm based on RMT are studied, and the influence of SNR, the length of sampled data and the number of receive antennas on performance of these algorithm is analyzed by the simulation results.(2) Based on RMT, a new PET algorithm(RPET) is proposed,significantly improving the correct estimation probability at low SNR and small sampled data size case, and performs slightly better than IWME.(3) Since the threshold predicted by using the ratio of confidence interval of adjacent random variables is a little high, RPET has relatively poor performace at low SNR case. Focusing on this shortcoming, we utilize the distribution characteristics of the mean square of eigenvalues of Wishart random matrix to design a new algorithm(WSE). WSE can accurately compute the decision threshold of test statistic by the distribution function of and accurate estimation of noise power. Simulation results show that WSE has better comprehensive performance than RPET, and significantly improve the correct estimation probability at low SNR and small sampled data size case. |
PDF Full Text Request