首页>
外国专利>
ATTENTION FEEDBACK BASED ROBUST SEGMENTATION OF ONLINE HANDWRITTEN WORDS
ATTENTION FEEDBACK BASED ROBUST SEGMENTATION OF ONLINE HANDWRITTEN WORDS
展开▼
机译：基于注意反馈的在线手写单词的鲁棒分割
展开▼
页面导航
摘要
著录项
相似文献
摘要
Embodiments of the present disclosure relate to a lexicon-free, attention-feedback segmentation approach for handwritten online Tamil words. Gross segmentation of the given word is performed by the Dominant Overlap Criterion Segmentation (DOCS) module into a set of stroke groups. Attention on certain spatial and temporal features detects segmentation errors, if any. The likelihoods feed back by the SVM as well as known statistics of features corrects the erroneous stroke groups to form valid symbols in the AFS module. The high success rate of segmentation holds promise in recognition of online handwritten words such as proper names and addresses, where it is not possible to invoke a finite lexicon. The reduction of the segmentation errors by the AFS module in turn leads to an improvement in the performance of the handwriting recognition system. Figure 2
展开▼
