Training-Based Multiple Source Tracking Using Manifold-Learning and Recursive Expectation-Maximization

Keyphrases

• Centroid 20%
• Comprehensive Simulation 20%
• Data-driven Approach 20%
• Exact Positions 20%
• Expectation Maximization 20%
• Feature Vector 20%
• Gaussian Mixture 20%
• Gaussian Parameters 20%
• Intermittency 20%
• Labeled Data 20%
• Large Database 20%
• Localization Accuracy 20%
• Localized Source 20%
• Manifold Learning 100%
• Mapping Function 40%
• Maximum Likelihood Estimation 20%
• Microphone Position 20%
• Mixture of Gaussian Model 20%
• Multiple Manifolds 20%
• Multiple Sources Localization 20%
• Multiple Speaker Tracking 20%
• Nonlinear Manifold 20%
• Number of Sources 20%
• Overlapping Speech 20%
• Partially Overlapped 20%
• Propagation Model 20%
• Recursive Expectation-maximization 100%
• Relative Transfer Function 100%
• Reverberant Enclosure 20%
• Reverberation 20%
• Semi-supervised Learning 20%
• Simple Propagation 20%
• Simulation Study 20%
• Source Position 40%
• Source Tracking 100%
• Sparsity 20%
• Speech Signal 40%
• Time of Use 20%
• Tracking Capability 20%
• Training Base 100%
• Training Stages 20%

Computer Science

• Baseline Method 20%
• Driven Approach 20%
• Feature Vector 20%
• Gaussian Model 20%
• Likelihood Estimation 20%
• Mapping Function 40%
• maximum-likelihood 20%
• Propagation Model 20%
• Semisupervised Learning 20%
• Simulation Study 20%
• Source Localization 20%
• Sparsity 20%
• Transfer Function 100%