Supervised neighborhood preserving embedding method based on kernel function

Abstract

The invention discloses a supervised neighborhood preserving embedding method based on a kernel function. According to the method, compared with a discriminant neighborhood embedding algorithm, the problem of unbalanced distribution of data samples can be processed, and the recognition rate is high; and training and test samples are transformed to a non-linear space through the kernel function, and a dimension reduction characteristic matrix is obtained through learning and training by employing category information of the known training samples so that the samples have better separability in the discrimination subspace.

Description

technical field

[0001] The invention belongs to the field of machine learning and pattern recognition, in particular to a supervised neighbor-preserving embedding method based on kernel functions. Background technique

[0002] Face recognition has attracted much attention due to its huge application prospects in public security, system files, and human-computer interaction. Face recognition is easily affected by many factors such as illumination, expression, posture, etc., and the higher the dimensionality of the image vector space, the more difficult it is to recognize. Effective feature selection and how to project the high-dimensional feature space into a suitable low-dimensional subspace have become important issues in the field of face recognition.

[0003] Neighborhood Preserving Embedding (Neighborhood Preserving Embedding, NPE) is a linear approximation algorithm for local linear embedding, which has the ability to preserve the local neighborhood structure informati...

Examples

Embodiment Construction

[0028] Below in conjunction with accompanying drawing, technical scheme of the present invention is described in further detail:

[0029] Such as figure 1 As shown, a supervised neighbor-preserving embedding method based on kernel function, including training and classification;

[0030] The training specifically includes the following steps:

[0031] Step 1, kernel function mapping:

[0032] Suppose the training sample set in the original space is Among them, c i is x i category label, c represents the number of categories, N represents the total number of training samples, and d represents the dimension of the training samples; the kernel function can explore the inherent geometric structure of the nonlinear space, through the function φ:∈R d →F maps the original d-dimensional data to a nonlinear feature space; where the function φ is K(x i ,x j )=i ),φ(x j )>;

[0033] Step 2, training data preprocessing:

[0034] For the spatial samples mapped by the kernel func...

technical field

[0001] The invention belongs to the field of machine learning and pattern recognition, in particular to a supervised neighbor-preserving embedding method based on kernel functions. Background technique

[0002] Face recognition has attracted much attention due to its huge application prospects in public security, system files, and human-computer interaction. Face recognition is easily affected by many factors such as illumination, expression, posture, etc., and the higher the dimensionality of the image vector space, the more difficult it is to recognize. Effective feature selection and how to project the high-dimensional feature space into a suitable low-dimensional subspace have become important issues in the field of face recognition.

[0003] Neighborhood Preserving Embedding (Neighborhood Preserving Embedding, NPE) is a linear approximation algorithm for local linear embedding, which has the ability to preserve the local neighborhood structure informati...

Examples

Embodiment Construction

[0028] Below in conjunction with accompanying drawing, technical scheme of the present invention is described in further detail:

[0029] Such as figure 1 As shown, a supervised neighbor-preserving embedding method based on kernel function, including training and classification;

[0030] The training specifically includes the following steps:

[0031] Step 1, kernel function mapping:

[0032] Suppose the training sample set in the original space is Among them, c i is x i category label, c represents the number of categories, N represents the total number of training samples, and d represents the dimension of the training samples; the kernel function can explore the inherent geometric structure of the nonlinear space, through the function φ:∈R d →F maps the original d-dimensional data to a nonlinear feature space; where the function φ is K(x i ,x j )=i ),φ(x j )>;

[0033] Step 2, training data preprocessing:

[0034] For the spatial samples mapped by the kernel func...

Images