Gesture recognition method based on Reders model

Abstract

The invention discloses a gesture recognition method based on a Reders model. The method specifically comprises the following steps: selecting a training data set, and taking an IPN Hand data set and a small amount of self-shot video data as the training data set; performing frame-by-frame hand posture estimation on the video in the training data set by using a MediaPipe framework, and extracting multi-frame gesture information; converting the extracted gesture information into time sequence data information by adopting an optical flow algorithm based on human body posture estimation; establishing a twin network, inputting the time sequence optical flow data obtained by processing, and performing training and optimization to obtain a pre-training model; and carrying out migration training on the pre-training model to obtain a final model. According to the method, a twin network architecture is adopted, the problem of small sample learning caused by incomplete data sets is solved, and meanwhile, a convenient, rapid and efficient solution is provided for adding new gestures for subsequent expansion.

Description

technical field [0001] The invention relates to the technical field of gesture recognition, in particular to a gesture recognition method based on the Reders model. Background technique [0002] In contemporary conference teaching and other scenarios, the control of presentation images, such as PPT, models, etc., mainly relies on physical media such as mouse and keyboard. Within the scope of the conference, it affects the freedom of conference presentation. Although there are some wireless presenters that can partially solve the above problems, the power of the controller and the inconvenient receiver limit the convenience of use; at the same time, the wireless controller cannot provide functions such as rotation and zooming, and lacks flexibility. This involves how to achieve better human-computer interaction. [0003] With the popularization of computer equipment, the important role of human-computer interaction in daily life is becoming more and more prominent. The deve...

AI Technical Summary

Problems solved by technology

[0007] The mainstream conference demonstration control system on the market focuses on realizing remote conference functions and central control functions, focusing on video, chat, recording, sharing, projection and other functions, ignoring the improvement of human-computer interaction in conference control functions

Images