MHW蒙古文脱机手写数据库及其应用

建立公开、权威的蒙古文手写数据库是研究和开发蒙古文手写识别系统的基础。该文在蒙古文编码、构词和语法的研究基础上,公开了一个蒙古文大词汇量脱机手写数据库MHW,其中训练集由5 000个单词构成,每个词采集了20个样本,共包含10万样本,测试集Ⅰ包含5 000样本,测试集Ⅱ包含14 085样本。该文利用蒙古文文字长度可变特征研究了自动错误检测算法,提高了字库的可靠性。在三种常用手写识别模型上评估了字库的性能,其中基于循环神经网络的模型表现出最佳性能,在字典受限条件下测试集Ⅰ的词错误率达到2.20%,测试集Ⅱ达到了5.55%。     

基于多尺度的蒙古文脱机手写识别方法

蒙古文的一大特点是字符无缝连接,因此一个蒙古文单词有多种字符划分方式。根据蒙古文这一特点,该文提出了多尺度蒙古文脱机手写识别方法,即让一个手写蒙古文单词图像对应多种目标序列,用多个目标序列同时约束训练模型,使得模型更加精准地学习手写图像的细节信息和蒙古文构词规则。该文提出了“十二字头”码、变形显现码和字素码3种字符划分方法,且拥有相互包含关系,即“十二字头”码可以分解为变形显现码、变形显现码可以进一步分解为字素码。多尺度模型首先用多层双向长短时记忆网络对序列化手写图像进行处理,之后加入第一层连接时序分类器做“十二字头”码序列的映射,然后是第二层连接时序分类器做变形显现码序列的映射,最后是第三层连接时序分类器做字素码序列的映射。用三个连接时序分类器损失函数的和作为模型的总损失函数。实验结果表明,该模型在公开的蒙古文脱机手写数据集MHW上表现出了最佳性能,在简单的最佳路径解码方式下,测试集Ⅰ上的单词识别准确率为66.22%、测试集Ⅱ上为63.97%。     

传统蒙古文与西里尔蒙古文相互转换方法的研究

传统蒙古文和西里尔蒙古文分别是在中国和蒙古国使用的蒙古文,它们的口语基本相同,但是书写形式完全不同。结合传统蒙古文和西里尔蒙古文的构词特点,提出了基于联合序列模型的传统蒙古文和西里尔蒙古文相互转换方法,并做了大量的相互转换实验。实验中,传统蒙古文到西里尔蒙古文转换系统的词误识率和字母误识率分别达到了18.38%和6.75%,西里尔蒙古文到传统蒙古文转换系统的词误识率字母误识率分别达到了18.77%和7.14%,基本达到了实用要求。     

连续字符识别的级联HMM训练算法

用较简洁的模型表达词典中成千上万的词条以及由这些词组成的类别无以计数的句子,是连续手写识别技术要解决的关键问题之一.利用HMM对时间序列的较强的建模能力这一特点,提出了可用于连续字符识别的HMM级联模型;给出了字符HMM模型连接概率和模型间状态转移概率的定义,并通过修正训练算法的重估公式,重估字符模型的连接参数,描述了用于手写体识别的字符HMM的设计方案;给出了级联训练算法重估公式和级联Baum-Welch训练算法描述.将级联方法应用于手写体英文单词的识别,结果表明,级联模型的描述能力与比较系统相比有明显的优势,获得了良好的识别性能.     

蒙古文停用词和英文停用词比较研究

该文采用联合熵算法(Union Entropy,UE)初步确定了蒙古文停用词,接着从初步确定的蒙古文停用词中去掉蒙古文实体名词及同形异义词,再通过对英文停用词和蒙古文停用词的词性比较,确定了蒙古文停用词表。最后用蒙古文停用词表和英文停用词表进行了文档信息检索的对比实验。实验结果表明,用该文所述方法确定的蒙古文停用词表进行蒙古文文档检索,比用英文停用词翻译成蒙古文进行蒙古文文档检索的准确率更高。     

基于分割识别的蒙古语语音关键词检测方法的研究

蒙古文属于黏着语,词根和后缀能够组合成近百万的蒙古文单词.现有的蒙古语大词汇量连续语音识别(LVCSR)系统的发音词典无法包含所有蒙古文单词.同时发音词典较大时,训练语料的稀疏将导致LVCSR系统的性能明显下降.为了解决LVCSR系统中大多数蒙古文单词的识别问题和蒙古语语音关键词检测系统中大量集外词的检测问题,结合蒙古文的构词特点,提出了基于分割识别的蒙古语LVCSR方法,并建立了对应的声学模型和语言模型.最后,将此方法应用到了蒙古语语音关键词检测系统中并在蒙古语语音语料上进行了测试.实验结果表明,基于分割识别的蒙古语LVCSR方法能解决大部分蒙古文单词的识别问题,并将蒙古语语音关键词检测系统的大量集外词转化成了集内词,大幅度提高了检测系统的查准率和召回率.     

基于词典、规则的斯拉夫蒙古文词切分系统的研究

斯拉夫蒙古文是蒙古国现行的文字,又称为西里尔蒙古文或新蒙古文。蒙古文词干和词缀包含着大量信息,斯拉夫蒙古文词切分是斯拉夫蒙古文信息处理众多后续工作的基础。该文尝试了将词典和规则结合的方法对斯拉夫蒙古文进行词切分。首先预处理部分蒙古文词,然后基于词典切分高频和部分不符合规则的词。最后对剩余的词,用切分规则生成多个候选的词切分方案,然后在这些方案中选出最优方案。通过两种方法的有机结合,发挥各自的优点,得到了性能较好的斯拉夫蒙古文词切分系统。     

基于规则和统计相结合的西里尔蒙古文到传统蒙古文转换方法

西里尔蒙古文与传统蒙古文分别是蒙古国与中国使用的蒙古文,西里尔蒙古文到传统蒙古文的转换工作不仅给两国同胞的交流带来更多的便利,而且对蒙古族的科学、文化和教育发展具有重要意义。本文结合规则与统计模型的优点,研究了西里尔蒙古文到传统蒙古文的转换方法。本文首先采用基于规则的方法对西里尔蒙古文集内词进行转换,其次对集外词的转换采用了基于联合序列模型的方法,并采用N-gram语言模型解决了一个西里尔蒙古文单词对应多个传统蒙古文单词的问题。实验结果表明,该系统单词转换错误率低至4.12%,基本达到了实用要求。     

蒙古文字母到音素转换方法的研究

针对蒙古文字母到音素的转换(grapheme to phoneme conversion, G2P)问题, 提出了基于规则的蒙古文G2P转换方法和基于联合序列模型的蒙古文G2P转换方法。实验结果表明, 利用联合序列模型的蒙古文G2P转换方法要明显好于基于规则的蒙古文G2P转换方法。并且建立的基于联合序列模型的蒙古文G2P转换系统的词误识率为1632%, 音素误识率仅为337%, 能达到实用要求。     

基于条件生成对抗网络的蒙古文字体风格迁移模型

蒙古文的每个字素在词的不同位置有着不同的书写形式,使得蒙古文字形结构多样且数量庞大,从而导致利用计算机辅助和传统人工方式设计蒙古文字体需要耗费大量的人力物力。故创建一种能自动生成蒙文字体风格的模型十分必要。国内外已有学者开展了对汉字和英文字体风格自动迁移的研究,但蒙古文领域仍处于空白阶段。因此,该文提出将条件生成对抗网络模型应用于蒙古文字体风格迁移,并给出了相关模型,实现了相应的算法和软件。在蒙古文字体数据集上进行实验,模型采用生成损失和判别损失衡量模型,Adam优化器自动调整学习率,逐渐减少差异值,直到生成器和判别器达到纳什平衡状态,可直接从蒙古文标题字体生成蒙古文手写体等字体,得到的生成字体样式基本接近真实字体样式,达到字体风格迁移的效果。     

Effective technique for the recognition of offline Arabic handwritten words using hidden Markov models

In this paper, we present a novel segmentation-free Arabic handwriting recognition system based on hidden Markov model (HMM). Two main contributions are introduced: a new technique for dividing the image into nonuniform horizontal segments to extract the features and a new technique for solving the problems of the skewing of characters by fusing multiple HMMs. Moreover, two enhancements are introduced: the pre-processing method and feature extraction using concavity space. The proposed system first pre-processes the input image by setting the thickness of the input word to three pixels and fixing the spacing between the different parts of the word. The input image is divided into constant number of nonuniform horizontal segments depending on the distribution of the foreground pixels. A set of robust features representing the gradient of the foreground pixels is extracted using sliding windows. The input image is decomposed into several images representing the vertical, horizontal, left diagonal and right diagonal edges in the image. A set of robust features representing the densities of the foreground pixels in the various edge images is extracted using sliding windows. The proposed system builds character HMM models and learns word HMM models using embedded training. Besides the vertical sliding window, two slanted sliding windows are used to extract the features. Three different HMMs are used: one for the vertical sliding window and two for the slanted windows. A fusion scheme is used to combine the three HMMs. The proposed system is very promising and outperforms all the other Arabic handwriting recognition systems reported in the literature.     

Confidence- and margin-based MMI/MPE discriminative training for off-line handwriting recognition

We present a novel confidence- and margin-based discriminative training approach for model adaptation of a hidden Markov model (HMM)-based handwriting recognition system to handle different handwriting styles and their variations. Most current approaches are maximum-likelihood (ML) trained HMM systems and try to adapt their models to different writing styles using writer adaptive training, unsupervised clustering, or additional writer-specific data. Here, discriminative training based on the maximum mutual information (MMI) and minimum phone error (MPE) criteria are used to train writer-independent handwriting models. For model adaptation during decoding, an unsupervised confidence-based discriminative training on a word and frame level within a two-pass decoding process is proposed. The proposed methods are evaluated for closed-vocabulary isolated handwritten word recognition on the IFN/ENIT Arabic handwriting database, where the word error rate is decreased by 33% relative compared to a ML trained baseline system. On the large-vocabulary line recognition task of the IAM English handwriting database, the word error rate is decreased by 25% relative.     

Maximum mutual information training for an online neural predictive handwritten word recognition system

In this paper, we present a hybrid online handwriting recognition system based on hidden Markov models (HMMs). It is devoted to word recognition using large vocabularies. An adaptive segmentation of words into letters is integrated with recognition, and is at the heart of the training phase. A word-model is a left-right HMM in which each state is a predictive multilayer perceptron that performs local regression on the drawing (i.e., the written word) relying on a context of observations. A discriminative training paradigm related to maximum mutual information is used, and its potential is shown on a database of 9,781 words. Received June 19, 2000 / Revised October 16, 2000     

HMM based online handwriting recognition

Hidden Markov model (HMM) based recognition of handwriting is now quite common, but the incorporation of HMM's into a complex stochastic language model for handwriting recognition is still in its infancy. We have taken advantage of developments in the speech processing field to build a more sophisticated handwriting recognition system. The pattern elements of the handwriting model are subcharacter stroke types modeled by HMMs. These HMMs are concatenated to form letter models, which are further embedded in a stochastic language model. In addition to better language modeling, we introduce new handwriting recognition features of various kinds. Some of these features have invariance properties, and some are segmental, covering a larger region of the input pattern. We have achieved a writer independent recognition rate of 94.5% on 3,823 unconstrained handwritten word samples from 18 writers covering a 32 word vocabulary