Feature face method is a face recognition method based on KL transform, which is an optimal orthogonal transform for image compression. After KL transform, a new set of orthogonal bases is obtained from a high-dimensional image space, and important orthogonal bases are retained, which can be transformed into a low-dimensional linear space. Assuming that the projections of human faces in these low-dimensional linear spaces are separable, these projections can be used as feature vectors for recognition, which is the basic idea of feature face method. These methods need more training samples and are based entirely on image gray level.

2. Face recognition method based on neural network.

The input of neural network can be the face image with reduced resolution, the autocorrelation function of local area, the second moment of local texture and so on. This method also needs more samples for training, but in many applications, the number of samples is very limited.

3. Face recognition method based on elastic graph matching.

Elastic graph matching method defines the invariant distance to common face deformation in two-dimensional space, and uses attribute topology diagram to represent face. Any vertex of the topological graph contains a feature vector, which is used to record the face information near the vertex position. This method combines gray features and geometric factors, and allows images to be elastically deformed during comparison, which has achieved good results in overcoming the influence of expression changes on recognition, and does not need to train multiple samples for a single person.

4. Face recognition method based on Hausdorff distance.

Psychological research shows that the speed and accuracy of recognizing contour maps (such as cartoons) are not worse than those of recognizing gray maps. LHD is based on the line graph extracted from the face gray image, which defines the distance between two line sets. The difference is that LHD does not establish a one-to-one correspondence between line segment sets, so it is more suitable for small changes between line segment diagrams. The experimental results show that LHD has different lighting conditions and different postures.

5. Face recognition method based on support vector machine (SVM)

Support vector machine (SVM) is a new hotspot in the field of statistical pattern recognition in recent years. It tries to make the learning machine reach a compromise on experience risk and generalization ability, so as to improve the performance of the learning machine. SVM mainly solves a binary classification problem, and its basic idea is to try to transform a low-dimensional linear inseparable problem into a high-dimensional linear separable problem. The usual experimental results show that SVM has a good recognition rate. However, it needs a large number of training samples (300 samples in each category), which is often unrealistic in practical application. Moreover, the training time of support vector machine is long, the method is complex, and there is no unified theory on how to get the kernel function.