Gesture recognition method based on image space pyramid bag of features

Abstract

The invention relates to a gesture recognition method based on an image space pyramid bag of features. The method comprises the following steps of carrying out gesture division of an image with a gesture; extracting and describing features of the divided gesture image; using extracted features to train a histogram intersection core support vector machine; obtaining a gesture type to which a feature vector of the gesture image belongs according to the support vector machine, in order to achieve gesture recognition. The method combines a space pyramid algorithm and a bag of features algorithm, and describes features of overall quantity and distribution of feature points of the gesture image. By means of the histogram intersection core support vector machine, classification of gesture features is achieved, and gesture recognition is achieved further. In addition, the rate of recognizing gestures of multiple similar types is improved.

Description

technical field [0001] The invention belongs to the field of image processing and pattern recognition, in particular to a gesture recognition method based on image feature extraction and support vector machine. Background technique [0002] Modern human-computer interaction is developing towards a more harmonious and natural direction. A hot issue in human-computer interaction research is to enable users to use computers in a convenient and natural way that humans are familiar with. Gesture language is simple and intuitive, and is an effective extension of human-computer interaction. It has a wide range of applications in smart home appliance control, robot control, sign language recognition, and computer game control. [0003] The key technology of gesture recognition lies in the two steps of feature extraction of gesture image and gesture recognition. Currently, the most commonly used image feature extraction methods include feature extraction using edge feature pixels, ...

AI Technical Summary

Problems solved by technology

[0006] (2) Using Fourier operators to describe gesture features and using the spatial distribution characteristics of palm fingers to represent gestures. These two methods solve the adaptability of features to scale, rotation, and illumination, but the amount of calculation is large and the calculation time is long.

[0012] (1) Template matching requires a large number of training images, and the calculation speed is slow

(2) The design of Adaboost classifier is complex, the amount of calculation is large, and the calculation speed is slow

However, the selected kernel function and the parameters of the kernel function have a great influence on the recognition results.

Images