Semi-supervised learning image recognition method based on convolution-stacked noise reduction coding network

Abstract

The present invention claims to protect a semi-supervised learning image recognition method based on convolution-stacked noise reduction encoding network, and proposes a semi-supervised learning network in combination with supervised training convolutional neural network and unsupervised learning stacked noise reduction autoencoder , to extract more expressive high-dimensional features when the sample size is insufficient, and train better network parameters. The convolutional neural network is used to perform supervised training on sample data to extract image features while avoiding the destruction of spatial features. Then input the feature vector obtained from the training to the stacked denoising autoencoder for unsupervised learning to achieve further feature learning and reduce the training difficulty caused by small samples, and then perform classification based on a multi-layer artificial neural network. The model is used to recognize complex characters, medium characters, simple characters, and similar characters in the HLC2000 handwritten Chinese character dataset for off-line handwritten Chinese character recognition. The experimental results show that the average recognition rate of the model can reach more than 97%, and the results are relatively stable.

Description

technical field [0001] The invention belongs to the technical field of feature extraction and recognition methods of off-line handwritten Chinese characters, in particular to a handwritten Chinese character image feature extraction and classification method based on a semi-supervised network. Background technique [0002] Due to its wide application prospects in handwritten character recognition systems such as manuscripts, bills, and certificate signatures, handwritten Chinese character recognition has received extensive attention and research since the last century. Handwritten Chinese character recognition is divided into two categories: online handwritten Chinese character recognition and offline handwritten Chinese character recognition. Among them, offline handwritten Chinese character recognition is more difficult to recognize due to noise interference during data collection. Commonly used handwritten Chinese character recognition methods include support vector machin...

AI Technical Summary

Problems solved by technology

Therefore, currently based on a single deep network training, it is impossible to fully extract the representative features of handwritten Chinese character images to achieve the optimal classification effect

Images