Inhalt

•  
   Contents
•  
   Introduction
•  
   Recognition, Reproduction, and Organization of Interaction Time Series
•  
   Research Objectives and Research Questions
•  
   Document Structure
•  
  Background: Machine Learning of Time Series
•  
   Machine Learning
•  
   Alignment Methods for Time Series Data
•  
   Dynamic Time Warping
•  
   Hidden Markov Models
•  
   Ordered Means Models
•  
  Classification
•  
   Bayes Classifier
•  
   Nearest Neighbor Classifier
•  
   Support Vector Machines
•  
  Clustering
•  
   K-means
•  
   K-trees
•  
  Application in Interaction Scenarios
•  
   Recognition
•  
   Reproduction
•  
   Organization
•  
  Performance Measures
•  
   Classification
•  
   Generalization Error
•  
   Receiver Operating Characteristic
•  
   Clustering
•  
   Summary
•  
  Theory and Definition of Ordered Means Models
•  
   Model Design
•  
   Sequence Length Distribution in OMMs
•  
   State Duration Probabilities in OMMs
•  
   Parameter Estimation
•  
   Efficient Computation of Production Likelihoods and Responsibilities
•  
   Influence of the Standard Deviation Parameter
•  
   Numerical Aspects
•  
   Computation in Log Space
•  
   Scaling of Probabilities
•  
   Relation to Hidden Markov Models
•  
   Summary
•  
  Experiments: Methods, Data Sets, and Application
•  
  Methods
•  
   Recognition
•  
   Organization
•  
   Prototype Property
•  
  Hyperparameter Selection
•  
   N-Fold Cross Validation
•  
   Initialization of OMMs and HMMs
•  
   Data Pre-Processing
•  
   Application in Experiments
•  
   Overview of Data Sets
•  
   Summary
•  
   Recognition of Interaction Time Series Data with OMMs
•  
  Classification with Ordered Means Models
•  
   Incremental Classification of Time Series with OMMs
•  
   Adaptive Learning
•  
  Low Latency Recognition of Interaction Time Series
•  
   Data Sets
•  
   Experiments
•  
   Results and Discussion
•  
  Robust Recognition: Incomplete Data
•  
   Data Set
•  
   Experiments
•  
   Results and Discussion
•  
   Robust Recognition: Influence of Transition Probabilities
•  
   Data Sets
•  
   Experiments
•  
   Results and Discussion
•  
  Scenario: Conversational Head Gestures
•  
   Data Set
•  
   Experiments
•  
   Results and Discussion
•  
  Scenario: Finger Pressure Patterns in Playing Musical Instruments
•  
   Data Set
•  
   Experiments
•  
   Results and Discussion
•  
   Incremental Recognition and Adaptive Learning
•  
   Data Set
•  
   Experiments
•  
   Results and Discussion
•  
   Summary
•  
  Reproduction of Interaction Time Series Data with OMMs
•  
   Definition: Prototype
•  
  Mouse Gestures
•  
   Data Set
•  
   Experiments
•  
   Results and Discussion
•  
  Scenario: Gesture Imitation for Virtual Agents
•  
   Data Set
•  
   Experiments
•  
   Results
•  
   Summary
•  
   Organization of Interaction Time Series Data with OMMs
•  
  Methods: Clustering of Time Series Data with OMMs
•  
   Partitioning Clustering: K-OMMs
•  
   Hierarchical Clustering: K-OMM-trees
•  
  Experiment: Partitioning of Time Series Data
•  
   Experiments
•  
   Results and Discussion
•  
   Experiments: Hierarchical Organization with K-OMM-trees
•  
   Experiments
•  
   Results and Discussion
•  
  Scenario: Gesture Recognition with K-OMM-trees
•  
   Data Set
•  
   Experiments
•  
   Results and Discussion
•  
   Summary
•  
  Conclusion
•  
   Contributions: Algorithms
•  
   Contributions: Empirical Data Analysis
•  
   Future Work
•  
   Expectation Maximization Algorithm
•  
  pyOMM: An OMM Package for the Python Programming Language
•  
   Installation Instructions
•  
   How to use pyOMM
•  
  pyKTree: A Python Implementation of the K-tree Algorithm
•  
   Installation Instructions
•  
   How to use pyKTrees
•  
   Bibliography