Quirks & Quarks

Forget face recognition — an AI can tell who you are by how you dance

A machine learning algorithm was able to identify individuals by their dance moves regardless of the music

A machine learning algorithm was able to identify individuals by their dance moves regardless of the music

Dancers wearing motion capture equipment moving to one of eight genres of music (Emily Carlson)

A group of researchers in Finland using an AI to study dance and music, made an accidental discovery recently.  Our dance moves are a distinctive and reliable way to identify us — at least for a computer.

Over the past few years, Emily Carlson and her colleagues at the Centre for Interdisciplinary Music Research at the University of Jyvaskyla in Finland have tried to better understand why music affects us the way it does. Carlson used motion capture technology to learn what your dance moves say about your personality. Motion capture, common in the film and video game industries, is a way to digitize movement so it can be studied or reproduced using a computer. 

Motion capture dancing (University of Jyvaskyla)

Dance like A.I. is watching

In their most recent study the team wanted to use machine learning to see if a computer could extract from digitized dance moves which genre of music the participants in the study were dancing to.  It didn't go well.

In the experiment, 73 participants were fitted with motion capture equipment. They then danced to eight different genres of music including blues, country, electronic, jazz, metal, pop, reggae and rap. 

Unfortunately the machine learning algorithm was able to identify the correct genre of music less than 30 per cent of the time. 

This graph show musical genre and accuracy per genre (University of Jyvaskyla)

You can dance, but you can't hide

The big surprise in this study was something Carlson was not looking for. Even though the computer could not identify the genre of music, it could identify the individuals dancing with 94 per cent accuracy, regardless of the genre of music.

The digitized movements apparently were distinctive enough for each individual that the machine was able to recognize the characteristic patterns of movement of each dancer. She suggests the machine is recognizing the kinds of cues that make it often possible for us to recognize individuals we know at a distance by the way they walk.

The only genres the machine leaning algorithm struggled to identify individuals dancing to were jazz and heavy metal,  possibly because, for metal in particular, the limited number of dance movements expressed such as the universal 'headbanging.'



To encourage thoughtful and respectful conversations, first and last names will appear with each submission to CBC/Radio-Canada's online communities (except in children and youth-oriented communities). Pseudonyms will no longer be permitted.

By submitting a comment, you accept that CBC has the right to reproduce and publish that comment in whole or in part, in any manner CBC chooses. Please note that CBC does not endorse the opinions expressed in comments. Comments on this story are moderated according to our Submission Guidelines. Comments are welcome while open. We reserve the right to close comments at any time.

Become a CBC Member

Join the conversation  Create account

Already have an account?