Error Correction in Handwriting Recognition with the Fusion of Pinyin Speech Recognition and Language Model

摘要：

It is proposed in this paper to correct the errors in Chinese handwriting recognition based on the fusion from other modalities, including Pinyin speech recognition and language model. Firstly, the scores derived in different modalities are mapped to the probabilities of being recognized correct using histograms. Then, different fusion algorithms, including ANN, linear and non-linear fusion algorithms, are used to re-organize the statistics from the different modalities to generate final Nbest lists. Experimental results demonstrate that the fusion could enhance the handwring recognition accuracy greatly.

关键词： Chinese handwriting recognition fusion Piny in speech recognition ANN N-best list

作者: Yinzi Chen Guohong Ding Yanming Zou Wendong Wang

作者单位: Beijing University of Posts and Telecommunications P.O. Box 79 Beijing 100876, China Nokia Research Center Beijing Laboratory Beijing 100176, China

会议类型: 国际会议

会议名称: 2010 International Conference on Circuit and Signal Processing(2010年电路与信号处理国际会议 ICCSP 2010)

会议地点: 上海

会议语种:英文

页码: 545-550

在线出版日期: 2010-12-25（万方平台首次上网日期，不代表论文的发表时间）