Vocal affect inference
Tal Sobol Shikler, Tomas Pfister & Peter Robinson
Recording multi-modal cues for HCI
The automated understanding and synthesis of natural expressions in speech is still a major challenge for human-computer interaction and in industries such the computer-generated animation.
The aim of this research is to advance the technology and the knowledge required for implementation of such systems. We have investigated how to infer the emotion and social information from naturally evoked expressions in speech, especially expressions in the context of natural HCI environments, nuanaces and temporal charchteristics of expressions. These can then be integrated into HCI systems and also be used to enhance systems for realistic synthesis.
In the context of HCI systems we are also exploring integration with multi-modal systems, including features that we take for granted in human communication such as the context and additional physiological cues.
- Visualizing dynamic features of expressions in speech
Analysis of affect in speech
- Affect editing in speech
Altering affect in speech
- Classification of complex information: Inference of co-occurring affective states from their expressions in speech
Journal paper describing the overall inference system
- Analysis of affective expression in speech
Tal Sobol Shikler’s PhD dissertation