Learning Outcomes to be assessed: • Basic knowledge of emotions models • Basic knowledge of how technology can be endowed with the ability to affectively and socially interact with its users • Understand the challenges that affective computing and/or HRI pose to the machine learning field • Identify the advantages and disadvantages of different approaches to addressing these challenges • Understand the ethical implication that affective technology may have on society
A part of the assessment of this module is through a group mini-project (software and report) on emotion recognition system and an individual report on the challenges that emotion recognition poses to the machine learning community based on the mini-project. The reports should be grounded on emotion theories and modelling approaches from the literature and provide a critical discussion of the approach used to tackle the project, of its limitations and of possible future solutions.
Mini-Project: You are asked to create a system to recognize human affective states/dimensions from one or two affective modality (e.g., facial expression, posture). This requires selecting or creating a database of emotion expressions for the selected modality and performing the necessary steps for creating and evaluating the emotion classifier. The project can be done in group. You are expected to submit well-commented code for the part of the software you implement. Here are the expected steps:
– Collect new data [10% of marks] or choose an existing database – Label/refine the labelling of the data (if necessary) [10% of marks] – Identify the discriminative features and critically discuss the process [30% of marks] – Build the recognition model [30% of marks] – Optimization process [30% of marks] – Model the fusion of multiple modalities [points 20% of marks] – Thorough evaluation of the models [points 20% of marks]