Finnish scientists learnt that their machine learning system is surprisingly good at identifying people by their dancing movements, after conducting an experiment on 73 volunteers.
In the article “Dance to your own drum: Identification of musical genre and individual dancer from motion capture using machine learning” published in Journal of New Music Research on December 20, Finnish scientists reported that their initial hypothesis was that genre would be a stronger identificator than personality traits.
This is not a new idea: the paper cites previous works with similar hypothesis, suggesting that dance movements strongly differ from one genre to another.
To test the relationship between musical genre and dance movements, scientists have applied a novel approach with capturing movements and creating an AI model to analyse the results.
As a result, scientists have found that the machine learning system can identify people with much more accurately – up to 90%. Genre classification, on the other hand, was not very successful:
“Against expectations, individual classification was notably more accurate than genre classification”
These findings have also provided additional insights, for example that movements of some parts of the body, like knees, could be a greater identificator, than other parts. One day it could lead to creating AI that can identify people by their dance with accuracy greater than 90%:
“notable individuality of movement patterns…should be explored with further research, for example by using stimulus manipulations other than genre, or considering individual differences at the level of personality or culture”