Robust Digital Processing of Speech Signals 1st Edition by Branko Kovacevic, Milan Milosavljevic, Mladen Veinovic, Milan Markovic ISBN 3319536133 973319536132

1. Speech Signal Modeling

1.1 Nature of Speech Signal
1.2 Linear Model of Speech Signal

2. Overview of Standard Methods

2.1 Autocorrelation Method
2.2 Covariant Method
2.3 Forward and Backward Prediction
2.4 Lattice Filter
2.5 Method of Minimization of Forward Prediction Error
2.6 Method of Minimization of Backward Prediction Error
2.7 Method of Geometric Mean
2.8 Method of Minimum
2.9 General Method
2.10 Method of Harmonic Mean
2.11 Lattice-Covariant LP Method
2.12 Basic Properties of Partial Correlation Coefficient
2.13 Equivalence of Discrete Model and Linear Prediction Model
2.14 Speech Synthesis Based on Linear Prediction Model

3. Fundamentals of Robust Parameter Estimation

3.1 Principles of Robust Parameter Estimation
3.2 Robust Estimation of Signal Amplitude
3.3 Fundamentals of Minimax Robust Estimation of Signal Amplitude
3.4 Recursive Minimax Robust Algorithms for Signal Amplitude Estimation
3.5 Statistical Models of Perturbations and Examples of Minimax Robust Estimator
3.6 Practical Aspects of Implementation of Robust Estimators
3.7 Robust Estimation of Parameters of Autoregressive Dynamic Signal Models
3.8 Non-recursive Minimax Robust Estimation Algorithms
3.9 Recursive Minimax Robust Estimation Algorithm
3.10 Fundamentals of Robust Identification of Speech Signal Model

Appendices
Appendix 1—Analysis of Asymptotic Properties of Non-recursive Minimax Robust Estimation of Signal
Appendix 2—Analysis of Asymptotic Properties of Recursive Minimax Robust Estimation of Signal Amplitude

4. Robust Non-recursive AR Analysis of Speech Signal

4.1 Robust Estimations of Parameters of Linear Regression Model
4.2 Non-recursive Robust Estimation Procedure: RBLP Method
    4.2.1 Newton Algorithm
    4.2.2 Dutter Algorithm
    4.2.3 Weighted Least Squares Algorithm
4.3 Comparison of Robust and Non-robust Estimation Algorithms
    4.3.1 Analysis of the Estimation Error Variance
    4.3.2 Analysis of Estimation Shift
4.4 Characteristics of M-Robust Estimation Procedure
    4.4.1 Model Validity
    4.4.2 Stability
    4.4.3 Computational Complexity
4.5 Experimental Analysis
    4.5.1 Test Signals Obtained by Filtering Train of Dirac Pulses
    4.5.2 Test Signals Obtained by Filtering of Glottal Excitation
    4.5.3 Natural Speech Signal
4.6 Discussion and Conclusion

5. Robust Recursive AR Analysis of Speech Signal

5.1 Linear Regression Model for Recursive Parameter Estimation
5.2 Application of M-Estimation Robust Procedure: RRLS Method
5.3 Robust Recursive Least-Squares Algorithm
5.4 Adaptive Robust Recursive Estimation Algorithm
5.5 Determination of Variable Forgetting Factor
    5.5.1 Approach Based on Discrimination Function
    5.5.2 Approach Based on Generalized Prediction Error
5.6 Experimental Analysis on Test Sinusoids
    5.6.1 Testing with Fixed Forgetting Factor
    5.6.2 Testing with Variable Forgetting Factor
    5.6.3 Testing with Contaminated Additive Gaussian Noise
5.7 Experimental Analysis of Speech Signals
    5.7.1 Test Signals Obtained by Filtering a Train of Dirac Pulses
    5.7.2 Test Signals Obtained by Filtering Glottal Excitation
    5.7.3 Natural Speech Signal
5.8 Discussion and Conclusion

6. Robust Estimation Based on Pattern Recognition

6.1 Unsupervised Learning
    6.1.1 General Clustering Algorithms
    6.1.2 Frame-Based Methods
    6.1.3 Quadratic Classifier with Sliding Training Set
6.2 Recursive Procedure Based on Pattern Recognition
6.3 Application of Bhattacharyya Distance
    6.3.1 Bhattacharyya Distance
6.4 Experimental Analysis
    6.4.1 Direct Evaluation
    6.4.2 Indirect Evaluation
6.5 Conclusion

7. Applications of Robust Estimators in Speech Signal Processing

7.1 Segmentation of Speech Signal
    7.1.1 Basics of Modified Generalized Maximum Likelihood Algorithm
    7.1.2 Robust Discriminant Function
    7.1.3 Tests with Real Speech Signal
    7.1.4 Appendix 4: Robust MGLR Algorithm (RMGLR)
7.2 Separation of Formant Trajectories
    7.2.1 Experimental Analysis
    7.2.1.1 Experiments with Synthesized Speech
    7.2.1.2 Experiments with Natural Speech
7.3 CELP Coder of Speech Signal
    7.3.1 LSP Parameters
    7.3.2 Distance Measure
    7.3.2.1 Cepstral Measure
    7.3.2.2 Likelihood Ratio
    7.3.2.3 COSH Measure
    7.3.3 Linear Prediction Methods with Sample Selection
7.3.4 Experimental Analysis