How about giving up mice and keyboards when interacting with a computer, using only gestures drawn in the air to access folders and files, as Tom Cruise did almost 20 years ago in the film Minority Report? Thanks to scientists at Sun Yat-sen University in China, this technology may be closer to becoming a reality.
The researchers developed a new algorithm capable of recognizing hand gestures with a good balance between simplicity, precision and efficiency. Unlike current methods, this system is not hampered by common limitations such as high computational complexity, low speed, or reduced number of recognizable gestures.
“One of the main features of the algorithm is its adaptability to different types of hands. The system first tries to classify variations between thin, normal and wide hands, based on three measurements that consider the relationships between palm width, length and finger size,” explains microelectronics professor Zhiyi Yu, lead author of the study. .
Want to catch up on the best tech news of the day? Access and subscribe to our new youtube channel, Kenyannews News. Every day a summary of the main news from the tech world for you!
manual recognition
After classifying all possible variations, the algorithm begins the gesture recognition process by making comparisons with samples of the same type of hand stored in the system. By using a gesture library with corresponding models, scientists improve the recognition rate without impairing the computer’s processing power.
The user’s hand type is paired with an adaptive database containing varied examples of human anatomy. To further improve accuracy, the gesture’s effective area ratio is calculated based on a lower bound rectangle, used to recognize preliminary movements and compare them to the stored information.
“Traditional simple algorithms tend to suffer from low recognition rates because they can’t handle different types of hands. This ends up making the system slow and imprecise, generating unsatisfactory results compared to other devices for interaction between machines and human beings”, adds Yu.
using shortcuts
The new method developed by the researchers uses so-called “shortcut features” to perform pre-recognition tasks. Although the algorithm can identify gestures among nine possible samples, comparing them to all the models stored in the system would be very time-consuming.
To solve this problem, scientists created a device that calculates the proportion of the area of the hands, selecting only the most likely gestures within the sample library. With that, the system manages to reduce the number of possible movements to just three, guaranteeing a much greater precision.
During the tests carried out in the laboratory, the results obtained showed that the new algorithm was able to identify hand gestures with an accuracy greater than 93%, even with the images being rotated, twisted or resized during the movement recognition process.
“The next step will be to improve the algorithm’s performance so that it can work in poor lighting conditions and with a greater number of gestures as a basis. This technology could pave the way for the development of new ways to control electronic equipment in the future”, concludes Professor Zhiyi Yu.
