Tom Troscianko, Alison Holmes, Majid Mirmehdi

1
Surveillance from all anglespredicting human
action

• Tom Troscianko, Alison Holmes, Majid Mirmehdi

2
Surveillance from all angles
There are lots of these things
3
Surveillance from all angles
But what do they allow us to do?
4
Our history

• It all started when we had a beautiful student
  dancer who asked Can we tell emotional state of
  the dancer from a point-light display?

5
(No Transcript)
6
Dittrich WH, Troscianko T, Lea SEG, Morgan D
(1996) Perception of emotion from dynamic
point-light displays represented in dance.
Perception 25, 727-738.

• Dynamic information is sufficient to allow
  assessment of emotional state of dancer

7
So we got an EPSRC grant

• With Majid Mirmehdi (Dept Computer Science,
  University of Bristol)
• Researcher Dr Alison Holmes
• Question is there information in the behaviour
  of individuals or groups that allows prediction
  of what happens next?

8
Feasibility study with CCTV clips

• 18 Incident clips, 10 sec each
• 18 Matches
• 64 Neutral
• 50 Experts, 50 novices
• Judging whether something bad happens, and what

9
Here is the kind of thing that we showed people

• Participants judged whether this is an incident
  or not
• We also asked them to indicate at what time they
  made their decision

10
(No Transcript)
11
Mean responses to this incident
12
Results for incidents
13
Results for matches
14
Troscianko T, Holmes A, Stillman J, Mirmehdi M,
Wright D, Wilson A (2004) What happens next? The
predictability of natural behaviour viewed
through CCTV cameras. Perception 33, 87-101

• Proof that the task is possible, though hard, for
  human observers

15
What next?

• Feasibility study suggested that people can do
  the task
• So the information is present in the CCTV images
• And we can identify the key actions (these are
  often the type of movement by individual or
  group)
• So we applied to EPSRC for a new grant
• But did not get it

16
What next?

• We still think that doing this study is
  worthwhile
• Need new partners in city councils, police, etc
• More generally, we believe that careful
  psychophysical experiments can inform computer
  science about where the information is, and what
  can be achieved