High accuracy handwritten Chinese character recognition using LDA-based compound distances

We consider the problem of similar Chinese character recognition in this paper. Engaging the Average Symmetric Uncertainty (ASU) criterion to measure the correlation between different image regions and the class label, we manage to detect the most critical regions for each pair of similar characters. These critical regions are proved to contain more discriminative information and hence can largely benefit the classification accuracy for similar characters. We conduct a series of experiments on the CASIA Chinese character data set. Experimental results show that our proposed method is superior to three competitive approaches in terms of both accuracy and efficiency.

/pdf/similar-handwritten-chinese-characters-recognition-by-16njdv7n5y.pdf

Similar Handwritten Chinese Characters Recognition by Critical Region Selection Based on Average Symmetric Uncertainty

In this paper, we propose a linear discriminant analysis (LDA)-based compound distance measure for discriminating similar characters in handwritten Chinese character recognition. The previous compound Mahalanobis function (CMF) is shown to be a special case of the proposed method. On finding similar character pairs by cross-validation using a baseline classifier, LDA is applied to each similar pair, and the LDA-based distance measure is combined with the discriminant function of the baseline classifier. In our experiments on the ETL9B and CASIA databases using the modified quadratic discriminant function (MQDF) as baseline classifier, the LDA-based compound distance is demonstrated to outperform the previous compound function methods.

/pdf/lda-based-compound-distance-for-handwritten-chinese-5c0365jni0.pdf

LDA-Based Compound Distance for Handwritten Chinese Character Recognition

Multi-Font Printed Tibetan OCR

Tibetan optical character recognition (OCR) system plays a crucial role in the Chinese multi-language information processing system. This paper proposed a new statistical method to perform multi-font printed Tibetan/English character recognition. A robust Tibetan character recognition kernel is elaborately designed. Incorporating with previous English character recognition techniques, the recognition accuracy on a test set containing 206,100 multi-font printed characters reaches 99.67%, which shows the validity of the proposed method.

New statistical method for multi-font printed Tibetan/English OCR

The complexity of pattern recognition increases along with the number of classes to be classified. As a result, large-scale pattern recognition problems such as Chinese character recognition are very difficult. On the other hand, with least number of classes, the research of pairwise classification has well developed theories and applicable methods. Thus, it is very practical to apply pairwise methods to solve multi-class and large-scale pattern recognition problems. In this paper, the multi-class problem has been broken down into a combination of pairwise problems so that multi-class problem can be solved through pairwise classifier combination. The method has been applied in Chinese character recognition. Comparing with primary system, the first candidate recognition rate increases from 89.25% to 97.94%; the error rate is cut by 80.84%.

Pairwise classifier combination and its application on Chinese character recognition

Due to the lack of research in the field of minority nationality character recognition, printing Tibetan character recognition was studied as an example in this paper. An experimental system was set up by applying fuzzy direction line features, which performs very effectively in the off-line Chinese character recognition, into printing Tibetan characters recognition. It can be seen from this primary study that this method is effective in printing Tibetan recognition.

Shaoping Ma

Papers

Pairwise classifier combination and its application on Chinese character recognition

A method of printing Tibetan character recognition