BayesianDecisionTheory_CaseStudies

BayesianDecisionTheory_CaseStudies - A bayesian approach to...

Info iconThis preview shows pages 1–5. Sign up to view the full content.

View Full Document Right Arrow Icon
Abayesian approach to human activity recognition (A. Madabhushi and J. Aggarwal, "A bayesian approach to human activity recognition", 2nd Inter- national Workshop on Visual Surveillance ,pp. 25-30, June 1999 (hard-copy)) Human activity recognition -Recognize human actions using visual information. -One of the hottest problems in computer vision (see reviewsection). -Applications include monitoring of human activity in department stores, air- ports, high-security buildings etc. -Building systems that can recognize anytype of action is a difficult and chal- lenging problem. Goal -Build a system which is capable of recognizing the following 10 (ten) actions (from a frontal or lateral view): (1) sitting down (2) standing up (3) bending down (4) getting up (5) hugging (6) squatting (7) rising from a squatting position (8) bending sideways (9) falling backward (10) walking -The frontal and lateral views of each action are modeled as individual action sequences. -Input sequences are matched against stored models of actions. -The input sequence is identified by finding the closest stored action.
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
-2- Approach -Human actions can be recognized by tracking various body parts. -People sit, stand, walk, bend down, and get up in a more or less similar fash- ion. -The head of a person movesinacharacteristic fashion during these actions. -The movement of the head of the subject ove rconsecutive frames is used to represent the above actions. -Recognition is formulated as Bayesian classification. Strengths and weaknesses -The system is able to recognize actions where the gait of the subject in the input sequence differs considerably from the training sequences. -I ti salso able to recognize actions for people of varying physical structure (i.e., tall, short, fat, thin etc.).
Background image of page 2
-3- Representation scheme of actions -Estimate the centroid of the head in each frame: ( x 1 , y 1 ) , ( x 2 , y 2 ) ,..., ( x n + 1 , y n + 1 ) -Find the absolute value differences in successive frames: X = ( dx 1 , dx 2 ,..., dx n ) Y = ( dy 1 ), dy 2 ,..., dy n ) where dx i = x i + 1 - x i and dy i = y i + 1 - y i -This representation scheme provides invariance . ... Head detection and tracking -The centroid of the head is tracked from frame to frame. -Accurate head detection and tracking are crucial. -Detection is done by hand in the current implementation.
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
-4- Bayesian formulation -Giv enaninput sequence, aposteriori probabilities are computed using each of the training actions. P
Background image of page 4
Image of page 5
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 17

BayesianDecisionTheory_CaseStudies - A bayesian approach to...

This preview shows document pages 1 - 5. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online