Inhalt

•  
   Acknowledgements
•  
   Contents
•  
  Introduction
•  
   Motivation and Goals
•  
   Multimodal Manual Interaction Data
•  
   Recognition of Manual Interaction Through Action Primitives
•  
   Structure of the Thesis
•  
  Conceptual Basis and Related Work
•  
   Action Primitive, Action, Activity
•  
   Unimodal and Multimodal Interaction Recognition
•  
   Neuroscientific and Psychological Experiments
•  
  Recognition of Interaction Through Decomposition
•  
   Change Point Detection for Recognition
•  
   State of the Art Segmentation Approaches
•  
   Recognition of Interaction with Multiple Modalities
•  
   Summary
•  
   Experimental Setup and Scenario
•  
   Scenario
•  
   Hardware Components
•  
   Hand Sensors
•  
   Object Sensor
•  
   External Setup View
•  
   Interaction Trigger
•  
   Overview of the Hardware Components
•  
   Ground Truth Acquisition
•  
   Manual Annotation
•  
   Automated Annotation using Audio Cues
•  
   Properties of Recorded Data
•  
   Mean and Variance Measures
•  
   Action-specific Variability
•  
   Inter- and Intrapersonal Variability
•  
   Constrained vs. Unconstrained Trials
•  
   Summary
•  
   Multimodal Interaction Decomposition: Theoretical Background
•  
  Decomposition Framework
•  
   Fearnhead's Algorithm
•  
  Modeling of Action Primitives
•  
   Linear Models
•  
   Constant Models
•  
   Threshold Models
•  
   Product Models
•  
   Mixture Models
•  
  Multimodal Bimanual Segmentation Approaches
•  
   Bimanual Segmentation Approach
•  
   Hierarchical Segmentation Approach
•  
   Parallel Segmentation Approach
•  
   Summary
•  
   Multimodal Interaction Decomposition: Experimental Results
•  
   Data Pool
•  
   Measures of Segmentation Quality
•  
   Segmentation Granularity Index
•  
   Temporal Accuracy Index
•  
   Segment Overlap Ratio
•  
   Missing Segments Index
•  
  Unimodal Segmentation
•  
   Parameter Overview and Evaluation Issues
•  
   Tactile Modality
•  
   Audio Modality
•  
   Joint-angles Modality
•  
   Comparison of Unimodal Segmentations
•  
   Bimodal Segmentation: Hierarchical Approach
•  
   Method and Model Overview
•  
   Segmentation of Constrained vs. Unconstrained Trials
•  
   Manual Annotation vs. Cue-based Ground Truth
•  
   Segmentation Evaluation for Three Human Demonstrators
•  
  Multimodal Segmentation: Parallel Approach
•  
   Method and Model Overview
•  
   Parameter Influence: Granularity and Modality Weighting
•  
   Parameter Influence: Constrained vs. Unconstrained Scenario
•  
   Segmentation of Constrained vs. Unconstrained Trials
•  
   Comparison of Parallel and Hierarchical Segmentation
•  
   Summary
•  
   Towards High-Level Modeling
•  
  Ordered Means Models
•  
   Means Vector and Emission Densities
•  
   Path Probabilities and Production Likelihood
•  
   State Duration Probabilities
•  
   Clustering with OMMs
•  
   Measures of Clustering Quality
•  
  Experimental Results
•  
   Data Pool
•  
   Parameter Estimation with Cross-validation
•  
   Clustering for Different Modality Combinations
•  
   Clustering of Fast and Slow Action Primitives
•  
   Summary
•  
   Conclusion and Outlook
•  
   Decomposition of Interaction
•  
   Higher-level Modeling
•  
   Final Comments
•  
  Instructions for Human Demonstrators
•  
   Action Execution: Unconstrained Scenario
•  
   Action Execution with Cues: Constrained Scenario
•  
  Annotation Rules
•  
   Tactile Tier
•  
   Audio Tier
•  
   Joint-angles Tier
•  
   Semantic Tier
•  
   Cues Tier
•  
  Unimodal Segmentation
•  
  Audio Modality
•  
   Autoregressive Model
•  
   Constant Model
•  
  Joint-angles Modality
•  
   Subsampling Rate s
•  
   Prior parameter
•  
   Bibliography