【作者】 刁红军；

【导师】 钱龙华；

【作者基本信息】 苏州大学 ， 计算机技术， 2009， 硕士

【摘要】 近几年来,智能手机应用软件和3G网络得到了迅速发展,手机已成为人们日常生活中必不可少的一部分。然而,目前在手机平台上实现词语级输入的拼音输入法还不是很多且功能也很有限。因此,研究针对手机平台的整句级汉字输入技术有着很强的科学研究和实际应用的价值。首先,本文提出了T10拼音编码方案,该方案能减少数字序列对应的拼音序列个数,并采用简化了的HMM(Hidden Markov Model)模型对数字拼音串做预处理,为整句输入做好了准备。本文根据现有基于N-gram的输入模型的不足之处,提出了基于HMM的N-gram智能输入模型。该模型对HMM的计算公式进行了如下改进:把观察事件的条件概率用条件概率函数代替,其值称为用户输入偏爱因子。由于输入过程中连续采用两次HMM模型,并且第一个HMM的输出作为第二个HMM的输入,本文把这种情况下的多个串联起来的HMM模型称为CHMM(Cascaded HiddenMarkov Model)模型。另外,还提出了输入法的一个自学习模型的结构。其次,为满足用户存储空间和运行时间性能上的要求,本文对基于CHMM模型的手机输入方案中所用文件格式进行了详细的分析和设计。采用2字节定长拼音数字化编号表示单音节拼音,简化了拼音码的存储格式和匹配算法。为使短语库文件与二元信息文件相关联,本文引入了短语ID的概念,并实现了短语到ID转换。为了尽量加快数据的查找与计算,本文使用了二分查找算法,以便提高系统效率。最后,本文提出了一个智能手机通用输入模型,并在Windows Mobile系统上应用该模型进行验证。更多还原

【Abstract】 As the rapid development of the smart phone application and 3G network in the recent years, Mobile phones have become an essential and integral part of our daily lives. However, there are few phrase-level Pinyin input methods able to achieve the requirement of Chinese character input on mobile phone platform. Therefore, it’s urgent to study the phrase-level Chinese character input technology on the mobile platform.Firstly, this thesis proposes a coding scheme, T10 pinyin, to reduce the number of the digital alphabet corresponding to the combination of a variety of Chinese phonetic alphabet. A simplified HMM model is used to preprocess the digital alphabet string for sentence input. Based on the inadequacies of existing N-gram model, a novel N-gram model for intelligent input which is based on HMM is proposed and it replaces conditional probability of observed events by the conditional probability function. The value of the conditional probability function is known as user preference factor. That model is called CHMM which uses two consecutive HMM model and takes the first one’s output as the second one’s input. Otherwise, a self-study model which can remember user’s input habit is presented.Secondly, to meet the requirements of user storage space and run-time performance, it analyzes and designs the storage structure of the data files used by the CHMM. A 2-byte fixed-length digital ID is used to represent Pinyin monosyllabic to simplify the storage format of phonetic code and correlative search arithmetic. To associate the phrase library file with the bi-gram file, the concept of the phrase ID is introduced to implement the transition from the phrase to the ID.Finally, a novel general input model for smart phones is proposed and its prototype based on Windows Mobile is developed to validate such model.更多还原