Multiuser Channel Estimation from Higher-Order Statistical Matrix Pencil
- PDF / 799,755 Bytes
- 10 Pages / 600 x 792 pts Page_size
- 84 Downloads / 217 Views
Multiuser Channel Estimation from Higher-Order Statistical Matrix Pencil Jing Liang Department of Electrical and Computer Engineering, University of California, Davis, CA 95616, USA Email: [email protected]
Zhi Ding Department of Electrical and Computer Engineering, University of California, Davis, CA 95616, USA Email: [email protected] Received 3 February 2002 and in revised form 11 August 2002 This paper presents a new statistical approach to the blind estimation of linear multiple-input multiple-output (MIMO) channels with finite impulse response. A matrix pencil is constructed from a set of fourth-order cumulant matrices of the channel output signals. The MIMO channel impulse responses can then be efficiently estimated from the generalized eigendecomposition of this cumulant matrix pencil. Random weighting is applied in the matrix pencil construction to improve the reliability of the algorithm. The proposed new method requires a relaxed channel identifiability condition and is robust in the sense that it does not require the exact knowledge of the MIMO channel order. Keywords and phrases: blind system identification, higher-order statistics, multiple-input multiple-output linear systems, matrix pencil, identifiability condition.
1.
INTRODUCTION
In recent years, blind estimation of multiple-input multipleoutput (MIMO) linear channels has become a well-known research problem in multichannel communications and signal recovery. Satisfactory solutions of this problem can find diverse applications in areas such as multiuser detection, array signal processing, speech processing, and multichannel biomedical signal recovery. The key objective of blind MIMO channel estimation is to determine the unknown matrix channel impulse response without direct training or knowledge of the channel input signals. The receiver must rely on the statistical information from the channel output signals. When the channel is a memoryless system, the problem is often known as blind source separation (or independent component analysis) with the goal of directly extracting source signals from the instantaneous mixtures without explicitly identifying the mixing matrix [1, 2, 3]. On the other hand, many multiuser systems must deal with dynamic channels that are characterized by a convolutive model. Once the dynamics of the MIMO system are estimated, techniques previously used in blind source separation of memoryless systems can be employed subsequently for individual signal separation.
Most existing approaches to blind channel estimation rely on the use of either second-order statistics (SOS) or higher-order statistics (HOS) of the channel output signals. As two equally important directions, second-order and higher-order methods have different characteristics suitable for different application scenarios. Compared with HOS methods, SOS methods may provide better performance for shorter data records [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18]. They also require a subsequent source separation step after identifying the channel
Data Loading...
