Researchers from China's Zhongshan University have developed a new gesture recognition algorithm. According to them, with its help it will be possible to control the computer with simple hand movements.
Optical gesture recognition is a laborious process due to its high computational complexity, low speed, and accuracy. However, the new development managed to strike a balance between all these aspects. The uniqueness of the technology lies in the definition of the type of human hand. The algorithm analyzes the user based on three dimensions: the width and length of the palm, as well as the length of the finger. In total, there are three types of hands: thin, normal and wide.
Subsequent gesture recognition will be based on the type of hand of a particular user. This allows you to significantly reduce the amount of required calculations and improve their accuracy. To further simplify the process, the algorithm, after determining the type of hand, reduces the recognition of the type of gesture to one of the three most likely of the nine supported. After that, a more complex and high-precision instrument is included in the work, which determines which specific gesture was shown.
When developing the technology, scientists used 40 volunteers who showed different gestures several times. Another forty people repeated them. The algorithm was able to recognize the displayed with an accuracy of 93%. It took into account that gestures could be rotated or scaled. In the future, scientists plan to improve the performance of the algorithm so that it can recognize gestures in low light or several at once.
0 Comments:
Post a Comment
Your comment and facebook share will be appreciated