Static sign language letter recognition system and method based on kinect sensor

Abstract

The present invention relates to the field of computer vision and intelligent human-computer interaction, in particular to a human-computer interaction system and an interaction method based on machine vision. The invention combines a Kinect sensor to propose a method for static sign language letter recognition based on an improved SURF algorithm. The Kinect sensor collects the depth image of the target area to segment the hand pixel area, which can overcome the interference caused by illumination changes and complex backgrounds; the improved SURF algorithm is used to extract feature points, and at the same time set the adaptive radius r, in the neighborhood with r as the radius By comparing the number of feature points and the distance between feature points, the SURF feature points are screened step by step, which not only greatly improves the recognition rate, but also ensures that the recognition work is effective in environmental factors such as skin color, lighting changes, complex backgrounds, and angle changes. Robustness to scale changes. In order to overcome the problem of high dimensionality of SURF feature vectors, the "one-to-one" classification method of SVM is used to classify and train SURF feature descriptors to obtain recognition results.

Description

technical field [0001] The invention relates to the field of computer vision and intelligent human-computer interaction, in particular to a human-computer interaction system and an interaction method based on machine vision. Background technique [0002] With the wide application of computers, human-computer interaction (Human Computer Interaction, HCI) has become an important part of people's daily life. The cognitive habits and forms that humans naturally communicate with the natural world are the development direction of human-computer interaction. Therefore, researchers are also working hard to make future terminals able to hear, see, speak, and feel. Simply put, human-computer interaction is the interaction between humans and computers. From keyboard to mouse control, then from voice to touch, and then to multi-touch, with the expansion of the human-computer interaction mode and the continuous penetration of non-professional people, the human-computer interaction mode...

AI Technical Summary

Problems solved by technology

However, in order to overcome the interference of complex background and illumination changes on the recognition work, this method adopts the depth image as the image to be detected, and the pixel value information of the depth image is converted from the distance value information. After the region segmentation results, the palm part is easy to calculate the wrong feature points due to the extremely similar pixel value information, which will interfere with the next recognition work

At the same time, the SURF eigenvector is 64-dimensional, and the eigenvector with too high dimension will also bring some interference to the recognition work.

Images