Home > Technology > Technical Articles > What are the identification method for face recognition system?
What are the identification method for face recognition system?
2014-08-21 09:48:22

1 based on the characteristics of the face (PCA) method for face recognition eigenface method of face recognition based on KL transform, KL transform is the image compression

A optimal orthogonal transformation. High-dimensional image space after the KL transform to obtain a new set of orthogonal basis, keep the orthogonal base important, from this

These radicals can turn into a low dimensional linear space. If the projection face in these low dimensional linear space is divisible, can put these projection

As the feature vector for recognition, the basic idea is the eigenface method. These methods need more training samples, and is entirely based on image

The statistical characteristics of the gray. At present, there are some improved eigenface method.

Face recognition method based on neural network 2 neural network input

Can reduce the resolution of the face images, the local autocorrelation function, local texture two moments. This method also need more

For training samples, while in many applications, the number of samples is very limited.

Face recognition method of elastic graph 3 elastic graph matching

Method defines a generally face deformation has a certain invariant distance in two-dimensional space, and the properties of topological graph to represent the people

Face, any vertex topology contains a feature vector, to near the vertex position information recording face. This method combines the gray.

And geometric factors, can allow the image has elastic deformation during the process of comparison, in overcoming the influence to the recognition of facial expressions and received good effect

Fruit, at the same time training for single person no longer need more samples.

4 line Hausdorff distance (LHD) method for face recognition

The psychology research indicate that, the human in the recognition profile (such as comic) speed and accuracy than any recognition of gray difference. LHD is based

To extract from the face gray image of line drawing, which is defined as between two line segment set distance, is not out of the ordinary, LHD

One one to establish the corresponding relation between different line segments between line segments, so it can adapt to small changes between the line graph. The experimental results show that, LHD

Have a very good performance in different illumination conditions and different attitude situation, but it is in the high expression recognition effect is not good.

5 support vector machine (SVM) method for face recognition in recent years, support vector machine is a new hotspot in the field of statistical pattern recognition, it tries to

Make learning to reach a compromise on the empirical risk and the generalization ability of machine, so as to improve the performance of learning machine. Support vector machine is mainly to solve the one

2 classification problem, the basic idea is to put a low dimensional linear inseparable problem into a higher dimensional linear separable problem.

The usual experimental results show that SVM has better recognition rate, but it needs a large number of training samples (each 300), which in practice to

It is not realistic to. And support vector machine training time is long, complex extraction method, kernel function has no unified theory.


  [Return Home] [Print] [Go Back]   

Technology

Contact Us

  • Contact Person:

    DongGuang HongZhao Innovation Electronic Co., Ltd.
  • Tel:

    0769-83532821
  • Fax:

    0769-86969800
  • E-mail:

    hongzhao@hongzhao-china.com