en-de
en-es
en-fr
en-sl
en
en-zh
0.25
0.5
0.75
1.25
1.5
1.75
2
S-SEER: A Multimodal Office Activity Recognition System with Selective Perception
Published on Feb 25, 20076223 Views
I will present the use of layered probabilistic representations for modeling the activities of people in a system named S-SEER. I will describe how we use the representation to do sensing, learning an
Related categories
Chapter list
S-Seer: A Selective Perception System for Multimodal Office Activity Recognition00:03
Overview of the Talk00:29
Background and Motivation01:40
Sensing in Multimodal Systems02:57
Seer Office Awareness System (ICMI 2002, CVIU 2004 (to appear))03:56
Multimodal Inputs04:37
HMMs for Behavior Recognition05:33
Several Limitations of HMMs for Multimodal Reasoning06:15
Seer Explored Layered HMMs (LHMMs) 07:47
Seer: Multi-Scale Activity Recognition08:23
TITLE09:08
SEER’s Architecture11:17
Value of LHMMs for Seer Task13:00
HMMs Inference in Seer14:14
LHMMs Inference in Seer15:43
Selective Perception Policies (ICMI’03) 16:59
Related work17:15
Policy 1: Expected Value of Information (EVI)17:39
Subsets of Features: Example18:35
Criticality of Utility Model18:57
Considering Outcome of Making an Observation19:58
Balancing Costs and Benefits20:41
Hypothetico-Deductive Cycle21:48
EVI with HMMs22:31
EVI in HMMs22:47
Policy 2: Heuristic Rate-based Perception22:59
Policy 3: Random Selection23:25
S-Seer23:29
Experiments with Selective Perception24:00
Comparison of the Selective Perception Policies24:07
Richer Utility and Cost Models24:55
Context-sensitive Cost Models25:22
Video26:06
Summary33:46
Future Work34:31
Thank you35:27