PREFACE
1 CHANNEL ESTIMATION AND EQUALIZATION USING HIGHER-ORDER STATISTICS
1.1 Introduction
1.2 Single-User Systems :Baud Rate Sampling
1.2.1 Cumulant Matching
1.2.2 Inverse Filter Criteria
1.2.3 Equation Error Formulations
1.2.4 Simulation Examples
1.3 Single-user Systems :Fractional Sampling
1.3.1 Cumulant Matching
1.3.2 Simulation Example
1.4 Multi-user Systems
1.4.1 Inverse Filter Criteria
1.4.2 Cumulant Matching
1.4.3 Simulation Examples
1.5 Concluding Remarks
Bibliography
2 PERFORMANCE BOUNDS FOR BLIND CHANNEL ESTIMATION
2.1 Introduction
2.2 Problem Statement and Preliminaries
2.2.1 The Blind Channel Identification Problem
2.2.2 Ambiguity Elimination
2.2.3 The Unconstrained FIM
2.2.4 Achievability of the CRB
2.3 CRB for Constrained Estimates
2.4 CRB for Estimates of Invariants
2.5 CRB for Projection Errors
2.6 Numerical Examples
2.7 Concluding Remarks
Appendix 2.A Proof of Proposition 2
Bibliography
3 SUBSPACE METHOD FOR BLIND IDENTIFICATION AND DECONVOLUTION
3.1 Introduction
3.2 Subspace Identification of SIMO Channels
3.2.1 Practical Considerations
3.2.2 Simplifications in the Two-Channel Case
3.3 Subspace Identification of MIMO Channels
3.3.1 Rational Spaces and Polynomial Bases
3.3.2 The Structure of the Left Nullspace of a Sylvester Matrix
3.3.3 The Subspace Method
3.3.4 Advanced Results
3.4 Applications to the Blind Channel Estimation of CDMA Systems
3.4.1 Model Structure
3.4.2 The Structured Subspace Method: The Uplink Case
3.4.3 The Structured Subspace Method: The Downlink Case
3.5 Undermodeled Channel Identification
3.5.1 Example: Identifying a Significant Part of a Channel
3.5.2 Determining the Effective Impulse Response Length
Appendix 3.A
3.A.1 Proof of Theorem 1
3.A.2 Proof of Proposition 3
3.A.3 Proof of Theorem 4
3.A.4 Proof of Proposition 5
Bibliography
4 BLIND IDENTIFICATION AND EQUALIZATION OF CHANNELS DRIVEN BY COLORED SIGNALS
4.1 Introduction
4.2 FIR MIMO Channel
4.2.1 Original Model
4.2.2 Slide-Window Formulation
4.2.3 Noise Variance and Number of Input Signals
4.3 Identifiability Using SOS
4.3.1 Identifiability Conditions
4.3.2 Some Facts of Polynomial Matrices
4.3.3 Proof of the Conditions
4.3.4 When the Input is White
4.4 Blind Identification via Decorrelation
4.4.1 The Principle of the BID
4.4.2 Constructing the Decorrelators
4.4.3 Removing the GCD of Polynomials
4.4.4 Identification of the SIMO Channels
4.5 Final Remarks
Bibliography
5 OPTIMUM SUBSPACE METHODS
5.1 Introduction
5.2 Data Model and Notations
5.2.1 Scalar Valued Communication Systems
5.2.2 Multi Channel Communication Systems
5.2.3 A Stacked System Model
5.2.4 Correlation Matrices
5.2.5 Statistical Assumptions
5.3 Subspace Ideas and Notations
5.3.1 Basic Notations
5.4 Parameterizations
5.4.1 A Noise Subspace Parameterization
5.4.2 Selection Matrices
5.5 Estimation Procedure
5.5.1 The Signal Subspace Parameterization
5.5.2 The Noise Subspace Parameterization
5.6 Statistical Analysis
5.6.1 The Residual Covariance Matrices
5.6.2 The Parameter Covariance Matrices
5.7 Relation to Direction Estimation
5.8 Further Results for the Noise Subspace Parameterization
5.8.1 The Results
5.8.2 The Approach
5.9 Simulation Examples
5.10 Conclusions
Appendix 5.A
Bibliography
6 LINEAR PREDICTIVE ALGORITHMS FOR BLIND MULTICHANNEL IDENTIFICATION
6.1 Introduction
6.2 Channel Identification Based on Second Order Statistics: Problem Formulation
6.3 Linear Prediction Algorithm for Channel Identification
6.4 Outer-Product Decomposition Algorithm
6.5 Multi-step Linear Prediction
6.6 Channel Estimation by Linear Smoothing (Not Predicting)
6.7 Channel Estimation by Constrained Output Energy Minimization
6.8 Discussion
6.8.1 Channel Conditions
6.8.2 Data Conditions
6.8.3 Noise Effect
6.9 Simulation Results
6.10 Summary
Bibliography
7 SEMI-BLIND METHODS FOR FIR MULTICHANNEL ESTIMATION
7.1 Introduction
7.1.1 Training Sequence Based Methods and Blind Methods
7.1.2 Semi-Blind Principle
7.2 Problem Formulation
7.3 Classification lf Semi-Blind Methods
7.4 Identifiability Conditions for Semi-Blind Channel Estimation
7.4.1 Identifiability Definition
7.4.2 TS Based Channel Identifiability
7.4.3 Identifiability in the Deterministic Model
7.4.4 Identifiability in the Gaussian Model
7.5 Performance Measure: Cramer-Rao Bounds
7.6 Performance Optimization Issues
7.7 Optimal Semi-Blind Methods
7.8 Blind DML
7.8.1 Denoised IQML (DIQML)
7.8.2 Pseudo Quadratic ML (PQML)
7.9 Three Suboptimal DML Based Semi-Blind Criteria
7.9.1 Split of the Data
7.9.2 Least Squares-DML
7.9.3 Alternating Quadratic DML (AQ-DML)
7.9.4 Weighted-Least-Squares-PQML (WLS-PQML)
7.9.5 Wimulations
7.10 Semi-Blind Criteria as a Combination of a Blind and a TS Based Criteria
7.10.1 Semi-Blind SRM Example
7.10.2 Subspace Fitting Example
7.11 Performance of Semi-Blind Quadratic Criteria
7.11.1 MU and MK infinite
7.11.2 MU infinite, MK finite
7.11.3 Optimally Weighted Quadratic Criteria
7.12 Gaussian Methods
7.13 Conclusion
Bibliography
8 A GEOMETRICAL APPROACH TO BLIND SIGNAL ESTIMATION
8.1 Introduction
8.2 Design Criteria for Blind Estimators
8.2.1 The Constant Modulus Receiver
8.2.2 The Shalvi-Weinstein Receiver
8.3 The Signal Space Property and Equivalent Cost Functions
8.3.1 The Signal Space Property of CM Receivers
8.3.2 The Signal Space Property of SW Receivers
8.3.3 Equivalent Cost Functions
8.4 Geometrical Analysis of SW Receivers: Global Characterization
8.4.1 The Noiseless Case
8.4.2 The Noisy Case
8.4.3 Domains of Attraction of SW Receivers
8.5 Geometrical Analysis of SW Receivers: Local Characterizations
8.5.1 Local Characterization
8.5.2 MSE of CM Receivers
8.6 Conclusion and Bibliography Notes
8.6.1 Bibliography Notes
Appendix 8.A Proof of Theorem 5
Bibliography
9 LINEAR PRECODING FOR ESTIMATION AND EQUALIZATION OF FREQUENCY-SELECTIVE CHANNELS
9.1 System Model
9.2 Unifying Filterbank Precoders
9.3 FIR-ZF Equalizers
9.4 Jointly Optimal Precoder and Decoder Design
9.4.1 Zero-order Model
9.4.2 MMSE/ZF Coding
9.4.3 MMSE Solution wit Constrained Average Power
9.4.4 Constrained Power Maximum Information Rate Design
9.4.5 Comparison Between Optimal Designs
9.4.6 Asymptotic Performance
9.4.7 Numerical Examples
9.5 Blind Symbol Recovery
9.5.1 Blind Channel Estimation
9.5.2 Comparison with Other Blind Techniques 324
9.5.3 Statistical Efficiency
9.6 Conclusion
Bibliography
10. BLIND CHANNEL IDENTIFIABILITY WITH AN ARBITRARY LINEAR PRECODR
10.1 Introduction
10.2 Basic Theory of Polynomial Equations
10.2.1 Definition of Generic
10.2.2 General Properties of Polynomial Maps
10.2.3 Generic and Non-Generic Points
10.2.4 Invertibility Criteria
10.3 Inherent Scale Ambiguity
10.4 Weak Identifiability and the CRB
10.5 Arbitrary Linear Precoders
10.6 Zero Prefix Precoders
10.7 Geometric Interpretation of Precoding
10.7.1 Linear Precoders
10.7.2 Zero Prefix Precoders
10.8 Filter Banks
10.8.1 Algebraic Analysis of Filter Banks
10.8.2 Spectral Analysis of Filter Banks
10.9 Ambiguity Resistant Precoders
10.10 Symbolic Methods
10.11 Conclusio
Bibliography
11 CURRENT APPROACHES TO BLIND DECISION FEEDBACK EQUALIZATION
11.1 Introduction
11.2 Notation
11.3 Data Model
11.4 Wiener Filtering
11.4.1 Unconstrained Length MMSE Receivers
11.4.2 Constrained Length MMSE Receivers
11.4.3 Example: Constrained Versus Unconstrained Length Wiener Receivers
11.5 Blind Tracking Algorithms
11.5.1 DD-DFE
11.5.2 CMA-DFE
11.5.3 Algorithmic and Structural Modifications
11.5.4 Summary of Blind Tracking Algorithms
11.6 DFE Initialization Strategies
11.6.1 Generic Strategy
11.6.2 Multistage Equalization
11.6.3 CMA-IIR Initialization
11.6.4 Local Stability of Adaptive IIR Equalizers
11.6.5 Summary of Blind Initialization Strategies
11.7 Conclusion
Appendix 11.A Spectral Factorization
Appendix 11.B CL-MMSE-DFE
Appendix 11.C DD-DFE Local Convergence
Appendix 11.D Adaptive IIR Algorithm Updates
Appendix 11.E CMA-AR Local Stability
Bibliography