MA-YOLO dynamic gesture rapid identification method based on double-path segmentation

Abstract

The invention provides a dynamic gesture rapid recognition method based on double-path segmentation MAYOLO, and the method comprises the steps: 101, proposing an MAYOLO algorithm, converting a YOLO backbone network into a lighter resnet34 network for feature extraction, adding an ASPP module to change a receptive field, adding an attention mechanism, and carrying out the autonomous learning better; 102, a complex environment gesture recognition technology based on double-channel segmentation is provided, and gesture recognition in a complex environment can be achieved by combining double-channel segmentation with a depth information segmentation image and a skin color segmentation image. Compared with a previous basic method, the method has the advantages that the precision is improved by5.4%, and the recognition speed is reduced to 50 ms or below.

Description

technical field [0001] The invention belongs to the field of image processing, and specifically designs a fast recognition method for MA-YOLO dynamic gestures based on two-way segmentation. Background technique [0002] With the development of science and technology, computers have penetrated into the lives of ordinary people and are playing an increasingly important role in various fields. The current scientific research field pays more and more attention to the interaction with computers. In recent years, the rapid development of virtual reality technology has greatly improved human research enthusiasm. Existing research mainly focuses on recognizing facial expressions, mouths, detecting head orientation, tracking gaze, recognizing gestures and positioning techniques, and interpreting human postures. Various gestures are widely used in scientific fields and daily life, and gesture recognition has become a key research topic at present. At the same time, the academic com...

AI Technical Summary

Problems solved by technology

[0003] However, because the hand itself is a complex part of the human body, it has the characteristics of spatial position differences, diversity, and complexity, and the incompatibility of the human body itself

Moreover, in the process of gesture recognition, the existing gesture recognition technology is greatly affected by external environment, background and other factors, and there is no way to reduce the influence of external factors in gesture recognition on the recognition process, which is a major bottleneck of gesture recognition

At the same time, the existing gesture recognition technology consumes too much time in the recognition process, and there is no way to complete the implementation of gesture instructions in real time. However, the implementation of instructions cannot be completed in real time. Meet the use in daily life

Images