1Project IdeasCSCI-534Affective ComputingMechanics• Small groups (3)– Identified roles• Proposal *was* due next week (Feb 8)• Proposal Presentation following week (Feb15)• Project Presentations (April 19 & April 26)2Recognition (voice)• Question: How do qualities of the voice revealaffect• Survey literature on recognizing affectivestates from voice signal– Psychology & Engineering• Implement a approach to processing thesignal– Perhaps leveraging existing software• Evaluate approach• ResourcesRecognition (face)• Question: How can we recognize emotionsfrom the face• Survey literature on recognizing affectivestates from face– Psychology & Engineering• Implement a si mple approach– Likely will need to leverage existing software forprocessing facial information (too difficultotherwise)• Evaluate approach in a game context• Requirements: (web) camera/ computer/faceprocessing software /image processingexpertise3Computational Modeling -e.g., Amygdala• Survey literature on Amygdala’s involvementin fear response/fear conditioning, includingcomputational work• Implement a simple computational model– Perhaps improving the LeDoux neural net model• Evaluate approach– E.g., contrast with LeDoux resultsModeling emotion’s impact onaction selection in agents• Survey literature on computational models of emotionin agent designs• Incorporate a simple model of emotion into existingsingle agent or multi-agent framework– POMDP, Decision Network, RAPs ?– Use a top-down, bottom-up or social theory• Evaluate approach– E.g., In some abstract scenario (eg bullying incident, bird)show impact on varying initial emotional state• Requirements– Expertise in, and perhaps access to software for someexisting agent system such as PsychSim or ABL4Expression (Body)• Question: How is emotion expressed by the body• Survey literature on the bodily expression of emotion– Psychology (e.g Walbott) and Engineering (e.g., Badler’s Emote)• Design a simple mapping from emotion to behavior animationmanipulation• Implement (partially)• Evaluate approach– Use classmates as subjects• Requirements– Expertise in graphics, and/or access to software for characterbehavior generationExpression (facial)• Question: How emotion is expressed facially andimpact of facial expression dynamics oninterpretation (including impact on recognition ofemotion, on social variables such as trust, etc.)• Survey literature of facial expressions and dynamics• Generate a set of facial stimuli– Varying the dynamics and timing of expression componentsin a 3d model of a head• Evaluate people’s interpretation– Use classmates/schoolmates as subjects or build a webstudy• Requirements: Expertise in Maya (we have a 3Dexpressive head to use)5Expression (dialog)• Question: How is emotion expressed in dialog (e.g.lexical choice)• Survey literature on dialog and expression of emotionin dialog• Propose a simple expressive NLG model (e.g. impacton lexical choice)• Possibly Implement a simplified version• Evaluate people’s interpretation– Use classmates/schoolmates as subjects or build a webstudyExpression (voice)• Question: How is affect expressed in voice (e.g.lexical choice)• Survey literature on dialog and expression of emotionin dialog• Propose a simple expressive NLG model (e.g. impacton lexical choice)• Possibly Implement a simplified version• Evaluate people’s interpretation– Use classmates/schoolmates as subjects or build a webstudy– Contrast to existing approaches (eg see
View Full Document