Research from MIT (3)
(Pentland, 2000)

'Smart Rooms' technology

Where people are:

2D tracking of user difficult (occlusion of data)
System augmented with a dynamic & constraint-based skeletal model
Sample rate of just 30Hz
Typical noise: 9mm for translation, 0.6 degrees for rotation

Pentland (2000)

Who people are:

Many face recognition systems highly reliable
Audio now employed to increase accuracy

Facial expression recognition:

Many useful applications: recognizing driver or student sleepiness
Lips an important part of facial expression
Head and lips are tracked
Hidden Markov Models used to statistically determine shape

Pentland (2000)