ActivitiesThe ISPLab (Image and Sound Processing Lab) has developed a strong expertise in multi-view analysis for 3D reconstruction (objects, faces, environments) and behavioural analysis (gesture classification and recognition): in particular, a large room, named "smart-space" is devoted to human action classification. A set of cameras is present at its walls, in a controlled environment, allowing real-time 3D reconstruction of a body model; gesture classification is performed through advanced pattern recognition techniques on these models, providing profitable tools to analyse real contexts like crowded places, remote assistance to old people and disabled persons or suspicious behaviours in sensible areas.
3D reconstruction techniques are also studied for face modelling and for developing recognition engines robust to different poses and illuminations. Further research topic in biometric field is iris segmentation and recognition from single image. Other topics studied at the ISPLab are: video coding (Scalable Video Coding, Distributed Source Coding); pattern recognition for video-surveillance (2D and 3D face recognition, object localization, tracking and classification for security applications); sound analysis (localization, separation, tracking and classification of sound sources); sound production (musical sound synthesis/processing) and processing (sound propagation modelling, real-time environmental acoustic rendering); in-solid vibrational analysis (touch localization/tracking for computer-human interaction).