A group of Chinese scientists led by Zhi Yu from Zhongshan University has developed a gesture recognition algorithm. It helps to control the computer with the help of hand movements.

Scientists noted that hand gestures are an important communication method that can be used to control a computer. Progress in camera design, image analysis, and machine learning is gradually making optical gesture recognition more convenient every year. But, modern technologies still have many limitations – high computational complexity, low speed, accuracy, or a small number of recognizable gestures. To solve these problems, Chinese scientists have developed a gesture recognition algorithm that strikes a balance between complexity, accuracy, and applicability.

One of the main features of the new algorithm is its adaptability to different types of hands. The algorithm first tries to classify the user’s hand type as “thin”, “normal” or “wide” based on three dimensions that take into account the relationship between palm width, palm length and finger length. If the classification is successful, in the following recognition steps only the input gesture is compared with the saved samples of the same hand type.

“Traditional simple algorithms tend to suffer from low recognition rates because they cannot handle different types of hands. Our algorithm first classifies the type of hand, and then, using gesture patterns from the libraries, it recognizes the required one, ”said Zhi Yu.

At the stage of preliminary recognition, the algorithm calculates the ratio of the area of ​​the hand to select the three most probable gestures out of a possible nine. This simple function is sufficient to narrow down the number of possible gestures to three, of which the final gesture is determined using much more sophisticated and highly accurate feature extraction. According to Zhi Yu, “The gesture pre-recognition stage not only reduces the amount of computation and hardware resources, but also improves the recognition speed without sacrificing accuracy.”

To develop the device, scientists took the participation of 40 volunteers who showed nine gestures several times, and another 40 volunteers to determine the accuracy of the system. The results showed that the algorithm can recognize hand gestures in real time with an accuracy of over 93%, even if the input gesture images were rotated or scaled. According to the researchers, in the future they are going to work on improving the performance of the algorithm in low light conditions and increasing the number of gestures.

For the first time, a way to control a computer with the help of hands was mentioned in the sci-fi blockbuster of 2002 “Minority Report” by the character of Tom Cruise, John Anderton, using special gloves, controls a transparent computer screen the size of a wall.