Automatic Recognition Method of Letter Images in English Self-Learning Based on Partial Differential Equation Method
Table 1
Characteristic description of letters.
Data characteristics
Feature description
Legendre moment
Before extracting features, simply normalize the image matrix. Based on BP neural network letter recognition, each sample is represented by a 121-dimensional feature vector.
Pseudo-Zernike moment
The preprocessing process is the same as Legendre, calculated to 9th order
Pseudo-Zernike moment
The preprocessing process is the same as that of Legendre. After calculating to the 8th order, a 36-dimensional feature vector is used to represent each sample.
Fourier transform
Extract from the upper left, upper right, lower left, and lower right of the character image matrix to obtain a 32-dimensional feature vector of the low-frequency region of the image matrix.
Primitive feature extraction
Combine each sample with 7 primitives to generate a 7-dimensional feature vector.
Edge feature extraction
Before extracting features, the image is refined into a skeleton image.
