Blind source separation

Title: Blind source separation in real time using Second Order statistics

A multi stage approach to speech enhancement methods improves the quality of separation over standard techniques such as spectral subtraction and beamforming. Two algorithms are implemented for convolutive mixtures in two of the important stages of a speech enhancement system, the source separation stage and the background denoising stage. For source separation, a blind source separation method based on second order statistics has been adopted whereas for background denoising…


1. Introduction
1.1 Scope
1.2 Organization of the Thesis
2. Background and Related work
2.1 Background
2.1.1. Instantaneous mixing model
2.1.2. Convolutive mixing model
2.2 Solution approaches for convolutive BSS
2 2.1 Extending ICA in time domain
2.2.2 Frequency domain BSS
2.2.3 Time-Frequency domain BSS
2.3 Related work on second order statistics
3. Foundations
3.1 Second order Statistics
3.2 Short Time Fourier Transform (STFT)
3.3 Steepest Decent Algorithm
3.4 Convolutive BSS
3.5 Cross-Correlation, circular and linear convolution
4. Offline BSS
4.1 Problem formulation
4.2 Derivation of Offline Algorithm
4.2.1 Cross-Correlation
4.2.2 Backward Model
4.2.3 Power normalization
4.3 Offline Algorithm
5. Online BSS
5.1 Introduction
5.2 Derivation in time domain
5.3 Frequency domain conversion
5.4 Power Normalization
6. Spectral Subtraction based on minimum statistics
6.1 Components
6.2 Algorithm
6.3 Subtraction Rule
7. Real Time Implementation
7.1 Requirements
7.2 Matlab implementation
7.3 Experiment Setup and algorithm
7.4 Simulation result
8. Evaluation and Results
8.1 Introduction
8.2 BSS evaluation method
8.3 Experiment and Results
8.4 Evaluation of Spectral subtraction method
9. Conclusion

Author: Lakmal Silva, Zhu Bo

Source: Blekinge Institute of Technology

Download URL 2: Visit Now

Leave a Comment