Feature extraction method for facial expression recognition based on Weber multi-directional descriptor

Abstract

The invention relates to a feature extraction method for facial expression recognition based on Weber multi-directional descriptors. Its main technical characteristics are: performing Gabor wavelet transformation on facial expression images, and merging Gabor features in all directions on the same scale; The feature image is divided into non-overlapping sub-blocks, and the graph structure is constructed in the horizontal, vertical, and two diagonal directions respectively; the eigenvalues ​​​​of the graph structure in the directions of 0°, 45°, 90°, and 135° are calculated, taking The largest of the four eigenvalues ​​is used as the differential excitation of the Weber multi-directional descriptor; the gradient directions of the central pixel in two mutually perpendicular directions are calculated separately, and the larger gradient direction of the two is used as the Weber multi-directional descriptor. main direction. The invention has a reasonable design, can extract more effective and more discriminative texture detail features, significantly improves the recognition rate of human facial expressions, and has better recognition stability and generalization ability, and can be widely used in human facial expressions recognition and other image processing fields.

Description

technical field [0001] The invention belongs to the field of image processing, in particular to a facial expression recognition feature extraction method (WOD-GS) based on Weber multi-directional descriptors. Background technique [0002] Facial expressions contain rich and complex emotional information and play an important role in human communication and interaction. In recent years, with the development of artificial intelligence, facial expression recognition has become a research hotspot in the field of affective computing. [0003] Feature extraction algorithms play a vital role in facial expression recognition systems. The existing facial expression feature extraction methods can be divided into four types: the first method is to use geometric features for feature extraction, and measure the position, distance, shape change and mutual ratio of significant changes such as eyes, eyebrows, and mouth. However, this method loses some important recognition and information...

AI Technical Summary

Problems solved by technology

The existing facial expression feature extraction methods can be divided into four types: the first method is to use geometric features for feature extraction, and measure the position, distance, shape change and mutual ratio of significant changes such as eyes, eyebrows, and mouth. Geometric features are used for expression recognition, but this method loses some important recognition and information, and the accuracy of the recognition results is not high

The second method is based on overall statistical features, mainly including Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), Locality Preserving Projection (LPP), etc. Using lower-dimensional or subspace information to calculate the similarity between images, because a large amount of detailed information is ignored, the interference of external factors leads to a serious decline in the recognition rate

The third method is based on frequency domain feature extraction, among which Gabor wavelet transform is a representative method. This method first transforms the spatial domain features of the image into frequency domain features, and then extracts relevant low-level features. However, after Gabor wavelet multi-scale and multi-directional transformation, the dimension of the feature matrix is ​​very high, and the time complexity of the feature extraction algorithm is very high, resulting in low recognition efficiency

The fourth method mainly uses the optical flow method and establishes an optical flow model to represent the structure and motion information of the face parts in the image. However, the calculation amount of this method is extremely large, which limits its further application.

In 2010, inspired by Weber's law, Chen Jie and others proposed Weber local descriptor (WLD). WLD uses two parts of local stimulus ratio and gradient direction to describe image texture details. However, its disadvantages are: poor calculation Only consider the contrast information between the central pixel and the surrounding pixels during excitation, ignoring the intrinsic relationship between the surrounding pixels

Later, some scholars improved the WLD algorithm from different angles, but these algorithms only calculate the gradient information in the horizontal and vertical directions, and the spatial structure information of the image is not fully utilized.

In 2011, Abusham et al. used the idea of ​​graph structure for feature extraction, and extracted image features by constructing a graph structure (LGS) in a 4×3 neighborhood. The disadvantage of this method is that the number of pixels on both sides of the target pixel is different. Algorithms are not symmetrical

In 2014, Mohd et al. proposed the Symmetric Graph Structure (SLGS) algorithm, which made up for the shortcomings of the original LGS algorithm to a certain extent. However, the above graph structure algorithm still lacks gradient direction information.

[0005] In summary, the existing local feature extraction algorithms for facial expressions still need to be improved in terms of recognition rate in characterizing facial features.

Images