Presentation for Paris Intelligence Artificielle Meetup #6
Summary:
Quick description of the process currently used at Footbar to start recognizing football gestures from only one acceleremoter attacher on the calf of the football player.
Please feel free to contact us if you're interested by our data or our project, we would be glad to hear from you.
4. WHAT DO WE WANT ?
• Classify the player’s activity at any given time knowing only its acceleration.
• First questions to answer before even thinking about machine learning :
How to define an activity / a gesture ?
How to slice (temporally speaking) our acceleration data to guess the right cla
right gesture ?
5. • Create a similarity with NLP field :
Slice the raw data into « elementary » windows, representing words.
Create sentences made of successive words, ie gestures.
Classify gestures from their included elementary windows.
OUR APPROACH
Footbar NLP
Elementary gesture Word
Player’s gesture /
activity (e.g. shot,
walk...)
Sentence
Player gestures’
classification
Sentence classification
6. OUR APPROACH
Process for characterizing a new report :
Slice it into elementary windows
Classify every elementary window
Slice the report into bigger windows (successive elementary windows)
Classify every slice
9. CREATE A DATABASE – FROM SCRATCH
Where to start ?
Easily recognizable gestures, at elementary level too : walk.
For every player : Annotate some walk at elementary level (and macro
level too)
Learn from these annotated elementary windows and classify others in
the signal.
Amongst newly predicted elementary windows, look for macro walk
pattern (crash – stance – takeoff – freefall).
When there’s a match, check on video that the player is walking.
Repeat this for other gestures.
10. ELEMENTARY GESTURE RECOGNITION
An elementary gesture = around 10 (200 ms) * 3 (axes) values of
acceleration.
Our data representation : 32 features (max, min, mean, std …).
Pros : Fast to compute, same length for all windows.
Cons : Axis orientation depends on the player.
Classificaton algorithm used : 2-nearest neighbors.
Pros : Fast to compute, works well on the same player’s data.
Cons : Can be a terrible predictor with new player’s data.
11. ELEMENTARY TO REAL GESTURE
A first approach : Bag of words modelisation of real gestures.
Example : someone walking.
Word Occurences
Crash 2
Stance 2
Takeoff 2
Freefall 2
12. ELEMENTARY TO REAL GESTURE
Real gesture classification :
Use bag of words representation as input of a classifier.
Adapt to different lengths of gestures (e.g. walk of 6s and walk of 2s).
Use relative frequency.
13. SO FAR
Elementary gestures :
> 30k gestures.
10 different elementary gestures.
97% of accuracy.
Player gestures :
> 3k gestures.
4 different gestures : rest, walk, run and shot.
14. PERSPECTIVES
So many things to explore…
Work around the features, the classification algorithms.
Recognize more gestures.
Add other player’s data for team interactions.
When more data : Look at deep learning possibilities.
Other topics to explore : Shot strength measurements, distances.
Long term : Player characterisation.