Researchers use computers to correctly identify individuals with near-perfect accuracy based on their dance movements to various genres of music.
JYVÄSKYLÄ, Finland — In cities like Beijing or London it’s hard to go anywhere without being filmed on camera. Unfortunately, if you were planning on dancing your way through such heavily surveilled areas to avoid detection, a new study finds you would actually be making yourself easier to identify. Researchers from the University of Jyväskylä in Finland say that every single person has their own unique way of dancing, and computers are able to ascertain the identity of dancers with startling accuracy.
According to the research team, regardless of the type of music, from jazz to reggae, the vast majority of people maintain a uniform uniqueness to their dancing style. It’s this ever present personality in each of our dance moves that makes it easy for computers to ID dancers.
Over the past few years, the study’s authors have been using the same motion capture technology used in Hollywood to analyze people’s dance moves and what they can tell us about the individual. Over the course of their research, they’ve noted that dance moves can provide a whole lot of information about a person; such as if they are extroverted, neurotic, what type of mood they happen to be in, and even how much they empathize with others.
Humorously, the research team hadn’t initially set out to use computers to identify dancers. The original plan was to use machine learning to determine the musical genre participants were dancing to at a particular moment.
“We actually weren’t looking for this result, as we set out to study something completely different,” explains first study author Dr. Emily Carlson in a release. “Our original idea was to see if we could use machine learning to identify which genre of music our participants were dancing to, based on their movements.”
In total, 73 dancers took part in the experiment. Each participant was motion captured as they danced to eight different genres: rap, reggae, blues, country, electronic dance, jazz, and metal. They were told to dance in whatever way felt natural.
“We think it’s important to study phenomena as they occur in the real world, which is why we employ a naturalistic research paradigm,” says Professor Petri Toiviainen, the senior author of the study.
Rather surprisingly, the machine learning algorithm actually wasn’t very good at identifying the musical genres, only offering a correct guess about 30% of the time. However, the computer was much better at identifying the dancers based on their movements. Among the 73 participants, the computer accurately determined who was dancing 94% of the time.
“It seems as though a person’s dance movements are a kind of fingerprint,” says Dr. Pasi Saari, another study co-author and data analyst. “Each person has a unique movement signature that stays the same no matter what kind of music is playing.”
Interestingly, researchers also noted that of all the studied musical genres, the computer had the hardest time identifying people who were dancing to metal.
“There is a strong cultural association between Metal and certain types of movement, like headbanging,” Dr. Carlson explains. “It’s probable that Metal caused more dancers to move in similar ways, making it harder to tell them apart.”
While these findings certainly have surveillance implications, the study’s authors say they are much more focused on what these results can tell us about how humans interact with music.
“We have a lot of new questions to ask, like whether our movement signatures stay the same across our lifespan, whether we can detect differences between cultures based on these movement signatures, and how well humans are able to recognize individuals from their dance movements compared to computers. Most research raises more questions than answers,” Dr. Carson concludes, “and this study is no exception.”
The study is published in The Journal of New Music Research.