Munier, Christian: Implementation and Evaluation of Acoustic Distance Measures for Syllables. 2011
Inhalt
- Abstract
- Motivation
- Introduction
- Automatic Speech Recognition
- Feature Extraction
- Acoustic Modeling with Hidden Markov Models
- Language Modeling
- Template Based Recognition
- Dynamic Time Warping
- Related Work
- Local Mahalanobis Distance in Template Based Speech Recognition
- Discriminative Locally Weighted Mahalanobis Distance in Template Based Speech Recognition
- Kullback-Leibler Divergence in Timbre Matching for Music Genre Classification
- Comparison of Model Parameters in Timbre Matching for Music Genre Classification
- Synopsis
- Requirements
- The Tutoring Scenario Revisited
- Conceptual Properties
- Properties Implied by Application in a Tutoring Scenario
- Discussion of Related Work Methods
- Architecture
- Dynamic Time Warping
- Local Distance Measures for Dynamic Time Warping
- Temporal Statistics
- Distance Measures for Temporal Statistics
- Used Software
- System Architecture
- Evaluation
- Prerequisites
- Evaluation Speech Corpus
- Selection of Syllables
- Estimation of Syllable Borders
- Statistical Models for Covariance Estimation
- Acoustic Features
- Methods
- Dynamic Time Warping with Mahalanobis Distance
- Mahalanobis Distance vs. Euclidean Distance
- Techniques for Covariance Estimation from a Gaussian Mixture Model
- Diagonal Covariance Matrices vs. Fully Occupied Covariance Matrices
- One Speaker vs. Arbitrary Speakers
- Automatic Segmentation vs. Annotated Segmentation
- Consideration of Acoustic Context
- Consideration of Dynamic Features
- Kullback-Leibler Divergence on Gaussian Models
- Comparison of Gaussian Model Parameters
- Synopsis
- Conclusion and Outlook
- Bibliography
- List of Figures
- List of Tables
- List of Requirements
- List of Desirable Properties