This Machine Learning Model can Identify you by Your Walking Style!


This AI system can identify people from their walking styles

Researchers gathered data from 127 different individuals, including 20,000 different footstep styles

It was very accurate with an error margin of 0.7%

Till date, the only ways to identify a person were either by their fingerprints or by CCTV footage. There are a few facial recognition methods but they have their own defects. So introducing a new machine learning algorithm that can identify people by their walking style.

This artificially intelligence system was developed by a group of researchers from the University of Manchester and the University of Madrid. The basic idea behind the study was to differentiate people based on their walking style and speed. According to the researchers, each individual has 24 different styles of movement and styles when they are walking and this was the basis of their model. This is equivalent to scanning their fingerprints or retina.

In order to train the AI, researchers had to feed a huge amount of training data to help it distinguish between different walking styles. So the team collected a database consisting of approximately 20,000 footstep signals from 127 different people. They used floor-only sensors and high resolution cameras to compile the database. This dataset is called SfootBD, which is used to develop the advanced computational models needed for automatic footprint biometric verification.

To test the final model, researchers put it in 3 different security scenarios. And the results were quite interesting, as it identified people with an error margin of 0.7%.

So Let’s Summarise:

This could be a great way to track criminals or people who are lost. Its applications are vast. Another interesting application of this: if someone was late to their flight and was running to catch it, this model could identify them and immediately alert the respective authorities.

Hope you liked the article! Have a nice day!

Leave a Reply

1 Comment threads
0 Thread replies
Most reacted comment
Hottest comment thread
1 Comment authors
koleksi vlogger terfavorit Recent comment authors
newest oldest most voted
Notify of