Research On Southeast-Asian Fax Screening And Search Based On Language Recognition

The research of this paper is a part of the ministry-authorized subject. The purpose is to meet the urgent need of some agents, to lessen the repeated work of officers and to improve the efficiency and correctness, with high-efficient processing of internet fax data.In this paper, basic principles and methods of language recognition are studied in depth, the extraction of statistical features and texture features, judgment algorithm, designing of classifier and the system of Southeast-Asian fax screening and search are expounded in detail.On the basis of the exhaustive research on the characteristics of Southeast-Asian language in the research of feature selection and extraction, the paper proposes an extraction method of superscript feature and feature of the right black pixel in the interconnected domain, improves the extraction methods of special characters feature. Meanwhile, the extraction method of multi-scale wavelet energy proportional feature is used. By experiments the features are proved effectively to reflect on the characteristics of languages studied in this paper.The judgment algorithm of classifier designing is studied to provide the high-efficient classifier designing with stable theory support, including linear discrimination function, minimum distance judgment method, nearest neighbor classifying method, non-linear judgment method, support vector machine and so on.Aimed at language recognition of five Southeast-Asian languages, eight languages including Southeast-Asian and three Asian languages based on Chinese, twelve languages including Southeast-Asian, Asian and European languages, the paper proposes and designs a hierarchical classifier based on the fusion of the multi-feature and multi-detection criterion in the system designing. The designed system of fax screening and search based on language recognition is proved to have a relative-satisfied screening correctness by testing on fax with different qualities.