LEARNING OUTCOMES

This course provides students with foundational conceptual knowledge, methodologies, and tools for designing, implementing, and evaluating computer systems that can capture, represent, and automatically analyze the behavior of their users (e.g., in terms of gesture, movement, facial expressions, speech) and interact with them by generating multisensory feedback (e.g., images, sounds, control of actuators) in real-time.

AIMS AND LEARNING OUTCOMES

The course aims at introducing the foundational knowledge needed for designing and developing computer systems that can interact with their users naturally, by exploiting multiple sensory channels. This requires students to know and apply technologies for capturing, representing, and automatically analyzing the behavior of the users – e.g., algorithms for detecting and analyzing gesture, full-body movement, facial expressions, and speech – and for generating multisensory feedback (e.g., images, sounds, and control of actuators) in real-time. At the end of the course, students will:

• Know and understand the motivations for using a multimodal interactive system for a specific application, the logical architectures that describe the major components a multimodal interactive system consists of, the guidelines for designing and developing multimodal interactive systems, the application areas where multimodal interactive systems can be successfully exploited.
• Know the most relevant devices for capturing data that can characterize the behavior of the users and understand how they work and when and how they can be used.
• Know the most important techniques for representing and automatically analyzing the behavior of the users and understand how and when to apply them. Techniques receive as input data coming from multiple sensor devices covering multiple sensory channels.
• Be able to analyze specific use cases in selected application areas to evaluate pros and cons for developing a multimodal interactive system rather than a traditional graphical user interface.
• Be able to design a multimodal interactive system and to implement its major components, by using the development tools presented during lectures and hands-on in the course.

PREREQUISITES

None. It is useful, even if not strictly required, having some basic knowledge on human-machine interaction topics (design, development, and evaluation cycle of traditional user interfaces, interaction design methodologies). It is also useful having a basic programming experience.

TEACHING METHODS

The course includes theoretical and practical lectures (approximately 32h of theoretical lectures and 16h of practical lectures for a total of 48h of lectures). Theoretical lectures aim at introducing the concepts and the techniques the course focuses on. Practical lectures consist of hands-on and enable students to apply the presented concepts and technologies in specific case studies. Students can attend the practical lectures by using their laptops in the classroom. Practical lectures exploit tools for gesture and movement analysis and for audio processing. Lectures will be in English.

SYLLABUS/CONTENT

1. Introduction to multimodal systems

• Post-WIMP and multimodal user interfaces 
• Definition of multimodal system 
• Motivations for developing multimodal systems
• Guidelines for designing multimodal systems
• Frameworks: the W3C Multimodal Interaction Framework, the multi-layered framework for analysis of nonverbal communication

2. Visual modality

• Video and motion capturing devices
• Techniques for automatic analysis of full-body movement and gesture: computation of movement features, segmentation of movement streams, gesture recognition, analysis of expressive content in full-body movement and gesture
• Techniques for automatic analysis of facial expression: face detection, computation of facial features, automatic extraction of Action Units
• Hands-on

3. Auditory modality

• Audio capturing devices
• Techniques for automatic computation and analysis of audio features: temporal, spectral, and cepstral features
• Introduction to automatic analysis of speech
• Hands-on

4. Multimodal fusion

• Levels of fusion: early fusion model, late fusion model
• Methods for multimodal fusion

5. Case studies

• Concrete examples of design and development of multimodal systems in selected application scenarios, including museums and cultural heritage, performing arts, education, well-being, and rehabilitation.

RECOMMENDED READING/BIBLIOGRAPHY

Learning material includes pdf copies of the slides presented at the lectures, examples and exercises for the practical lectures, and a collection of scientific papers the teacher provides the students with. Learning material is made available on AulaWeb. Learning material is provided in English.​