Blind Estimation Of The Number Of Transmit Antennas Based On Random Matrix Theory

Next-generation wireless communication systems will provide higher transmissionrates and better quality of service, in the context of spectrum resource constraints,multi-input multi-output (MIMO) technology can provide technical support for therealization of this goal. However, with the development of MIMO, a series of newchallenges appear for the estimation and identification in blind identification scene, suchas blind estimation of the number of transmit antennas(BENT), blind codingidentification. As a prerequisite for the follow-up blind channel estimation, blind codingidentification, blind signal demodulation and other key technologies, the accuracy ofBENT has a significant impact on the correct identification of blind MIMO systems.Focusing on the study of BENT algorithms based on random matrix theory(RMT), thefollowing tasks are made in this thesis.(1) Three typical BENT algorithms, which are respectively based on informationtheoretic criteria (including AIC and MDL), predicted eigenvalues threshold(PET) andGerschgorin disk, are introduced. Moreover, these algorithms are analyzed bytheoretical and simulation analysis.(2) A novel BENT approach based on the ratio of the maximum eigenvalue tominimum eigenvalue (MME) is introduced from spectrum sensing. The simulationresults show that MME can obtain a performance close to PET.(3) Since the MME algorithm utilizes the limit of the minimum eigenvalue tocalculate the decision threshold, resulting in a relatively poor performance in a smallnumber of samples. In response to this shortcoming, we propose an improved methodwhich only take advantage of statistical characteristics of the maximum eigenvalue (ME)to decide the threshold. Compared with MME, ME can effectively improve theperformance in low SNR and small number of samples.(4)Moreover, as a reference for future research, two improvement ideas of MEalgorithm are proposed: studying the non-asymptotic distribution of RMT andconstructing self-adaptive decision threshold.