During the past two years, the MUCATAR team successfully developed
algorithms and implemented software for single-person robust tracking,
multi-people tracking, and joint tracking and activity recognition, in a
Sequential Monte Carlo (particle filtering) framework. Most of the work has
been implemented and/or evaluated in the meeting room scenario. The main
achievments are:

- the development of novel data fusion mechanisms to improve tracking, via
   - visual cue fusion (shape/color with motion) for stable tracking [1].
   - audio-visual fusion for speaker tracking with one or more cameras [2,3].

- the development of mixed-state models (which combine continuous
  motion parameters and discrete labels in a joint distribution) for
  tracking and recognition, including
   - multi-camera speaker tracking (the discrete variable represents the
     specific camera view in which the speaker appears) [3].
   - joint head tracking and head pose estimation (the discrete
     variable denotes a head pose/appearance examplar) [4].

- the development of new sampling methods for improving tracking efficiency
  and performance, including
   - the use of motion as proposal distribution for single-object tracking [5],
   - a distributed partitioned sampling strategy for multi-object tracking [6]

- data collection and development of performance evaluation procedures, including
   - a data set for the evaluation of multi-people and AV speaker
     tracking algorithms with precise 3-D groundtruth.
   - a data set for the evaluation of joint head tracking and head pose recognition.
   - a protocol for performance evaluation of multi-object tracking algorithms.

Current work focuses on

- the extension of multiple-people tracking algorithms (multi-camera, visual
  and audio-visual).
- the development of algorithms combining HMMs and particle filters to recognize,
  while tracking, more precise and complex activities (e.g. head and body gestures).
- the definition with other IM2.SA partners of a common data set for evaluation
  of multiple people tracking algorithms in surveillance scenarios.