自动发音错误检测中基于最大化F1值准则的区分性特征补偿训练算法

为提高自动发音错误检测性能,提出一种区分性特征补偿训练算法.该方法将高斯后验概率矢量经过线性变换后作为偏移量补偿至传统的谱特征.将经过正确度标注的语音数据库上的发音错误检测F1值的最大化作为变换参数的训练准则.推导了目标函数对变换参数的偏导数公式,并利用无约束参数优化例程L-BFGS更新变换参数.发音错误检测实验表明该方法能够有效增大训练和测试集的F1值.并且训练和测试集的精确度、召回率也都有明显提高.在特征优化的基础上进行模型参数训练,检错性能较单独的区分性特征训练、单独的区分性模型训练都有进一步改进.     

一种基于检错音网络的发音错误检测新算法

该文通过将计算机辅助语言学习(Computer Assisted Language Learning, CALL)系统的标准发音网络加入插入和删除路径的方法提出了一种发音错误检测新算法：检错音网络(Error-Detecting Network of Pronunciation, EDNP)错误检测算法。该算法首先对待测语音进行EDNP错误检测,然后通过对检错音网络的二级识别结果使用一级多候选词图进行错误召回的策略,进行错误检测,该算法易于实现并具有平台无关性。实验结果表明：该文提出的EDNP错误检测方法在中国四级考生语音测试库上使删除错误虚警率和漏报率分别达到7.38%和12.25%,插入错误虚警率和漏报率分别达到4.94%和26.17%,且客观评分与专家评分相关度比强制对齐方法的相关度提高了4.29%。     

英文发音中错误语音自动识别系统设计

传统的英文发音识别系统对于学习者的错误发音不能及时进行反馈与纠正,存在误导学习者以及学习者英文水平提高缓慢的弊端。在此设计新的英文发音错误语音自动识别系统,其由语音录制模块、语音播放模块、英语发音评分模块和发音共振峰图像显示模块构成,给出评分模块的发音评分流程,实现英文发音的有效评分以及评分的存储,系统通过发音共振峰图形显示模块,清晰地表达出学习者发音与标准发音的不同之处,纠正其错误读音。通过英语音素检错程序使用独立阈值的方式来提高错误读音的检测性能,对不同音素用独立阈值进行衡量,使得英语发音中的错误语音自动识别结果更加科学化、精准化。实验结果表明,所设计的系统具有较高的错误语音自动识别能力。     

声韵母约束扩展识别网络的发音偏误检测

发音偏误检测是计算机辅助发音训练（Computer Aided Pronunciation Training ,CAPT）的重要组成部分。为了在机器辅助语料标注任务或者缺少标注语料的偏误检测任务上提高性能,本文提出解码时使用声韵母约束的扩展识别网络方法。该方法将传统的语音识别中解码的自由文法循环（free grammar loop）部分换成结合声韵母交替以及字数限制规则的扩展识别网络,可以对全音素进行偏误检测, 并且不会出现插入删除错误。相比于传统的扩展识别网络,这种约束的扩展识别网络不需要大量的语料标注和分析。相对于传统的发音良好度评价方法（Goodness of Pronunciation, GOP）, 基于这种拓展识别网络的方法不仅可以对二语学习者的发音进行正误的检测,还能给出具体的错误反馈。实验结果表明,本文提出的基于声韵母约束拓展识别网络的方法在挑错任务上优于传统的发音质量评估（GOP）的方法,其错误接受率为29.2%,错误拒绝率为22.9%,诊断准确率为76.6%。比GOP方法的诊断准确率相对提升15.5%,并且模型相较于无标注经验汉语母语者能检测出更多偏误。      

基于语音识别的英语口语发音校正系统

针对传统英语口语发音校正系统对学生发音错误不能及时校正、反馈等问题，设计一种基于语音识别的英语口语发音校正系统。该系统由口语发音特征提取与校正两部分构成。校正是整个系统核心部分，通过标注结果做灭错计算，识别出错误音频；根据反馈路径、序列方式得出音频对照结果，利用鲁化方式进阶控制对照结果，完成口语发音校正。实验结果表明，所提方法可以将错误发音识别并校正，同时该系统响应时长较短，证实系统性能良好，实现了英语口语发音校正系统的科学化、精准化。     

一种改进的GOP算法在区分性训练的应用

自动发音错误检错中基于最大化F1值的区分性训练方法是最近提出来的一种声学模型训练方法,该方法能够有效增大发音检错系统中的训练和测试数据检错的Fl值。对发音质量评估方法上进行研究,提出一种改进的GOP算法来替代传统的GOP算法,改进GOP算法把传统地GOP算法的先求后验概率再求时间归一化改变成先求时间归一化再求后验概率。根据改进GOP算法给出了使用改进GOP算法最大F1准则的参数更新公式,发音检错实验结果表明基于改进的GOP算法的最大F1值准则训练较使用传统的GOP算法具有过训练抑制性好,在训练机上较低的目标函数值上能达到较高的测试集上的F1值等较好的性能。     

基于HMM／VQ的认人的中等词表连续语音识别

本文讨论基于隐马尔可夫模型(HMM)和矢量量化(VQ)的连续语音识别方法。用这种方法,对每个单词作成一个HMM,对多个模型组合成的状态转移网络搜索其状态转移的最佳路径,从而实现不预先进行单词切分的连续语音的识别,使用有限态文法约束及其它一些改善识别性能的措施,演示系统能识别特定人的18种英语句式,150个单词,用312个话句(共有2710个单词)进行测试,识别延迟时间为发音时长的62％,发音速度平均为每秒2.32个单词,单词识准率为97.3％。     

低数据资源条件下基于结构信息共享的无切分维文文档识别字符建模

无切分维吾尔文文档识别技术能够有效避免字符切分错误,但是对于低数据资源的新样本类型,原有模型往往难以获得较高的识别性能。为此,该文提出共享常用维文字体间相对稳定的字符结构信息,并用Bootstrap方法提高样本利用效率的解决方法。通过在实际书籍样本上的实验表明,仅利用规模约原始训练样本1/5的新类型样本,该方法在测试集上的平均字符识别准确率就可以达到95.05%;而与常用的最大后验概率估计方法相比,也能使识别错误率相对降低55.76%~63.84%。因此,该方法能够有效解决低数据资源条件下的维文字符建模问题,实现对新样本类型的高性能识别。     

英文语音发音标准化的模式识别对比方法改进分析

针对英文语音发音标准化评价准确性不高的问题,提出一种基于发音特征倒谱系数感知的英文语音发音标准化的模式识别对比方法。首先构建英文语音发音的语音信号采集模型,对采集的英文语音信号进行发音器官的动作属性配对描述。然后提取英文语音发音信号的倒谱特性,采用梅尔频率倒谱系数感知方法进行英文语音发音特征建模和发音位置及方式的模式识别,为语音发音提供标准化对比模式。最后进行实验分析,测试结果表明,采用该方法进行英文语音发音特征检测和模式识别的准确度较高,对发音特征的声学建模有效可靠。     

局部轨迹信息的联机手写维吾尔文单词过切分

将手写单词分割成字符是一项艰巨的任务.本文提出一种基于局部轨迹信息的单词过切分方法.该方法中,在不预先对附加笔划进行分组,不确定基线位置和不做倾斜校正的情况下探索单词轨迹的局部信息进行过切分.利用的局部信息包括手写轨迹中的平直点、局部最大和最小、局部最右和最左.探索手写单词原轨迹点序列中的局部最大值和局部最小值的点,从中找出切分点.初步确定的切分块里找到最右边和最左边的点更新切分点.本文提出的方法在5个不同作者的500个无约束手写单词样本上进行测试.平均字符切分召回率和正确率分别为93.35％和69.97％.     

Multicriteria‐Based Computer‐Aided Pronunciation Quality Evaluation of Sentences

The problem of the sentence‐based pronunciation evaluation task is defined in the context of subjective criteria. Three subjective criteria (that is, the minimum subjective word score, the mean subjective word score, and first impression) are proposed and modeled with the combination of word‐based assessment. Then, the subjective criteria are approximated with objective sentence pronunciation scores obtained with the combination of word‐based metrics. No a priori studies of common mistakes are required, and class‐based language models are used to incorporate incorrect and correct pronunciations. Incorrect pronunciations are automatically incorporated by making use of a competitive lexicon and the phonetic rules of students' mother and target languages. This procedure is applicable to any second language learning context, and subjective‐objective sentence score correlations greater than or equal to 0.5 can be achieved when the proposed sentence‐based pronunciation criteria are approximated with combinations of word‐based scores. Finally, the subjective‐objective sentence score correlations reported here are very comparable with those published elsewhere resulting from methods that require a priori studies of pronunciation errors.     

基于语音识别的发音学习技术

在语言发音学习中,有效的反馈对学习者有很大的帮助。计算机辅助发音学习系统可以给学习者有效的发音指导。就目前基于语音识别的发音学习技术进行介绍,给出系统原理框图,对一些关键技术和问题进行探讨,并对其发展进行展望。     

Identification of articulation error patterns using a novel dependence network

Articulation errors seriously reduce speech intelligibility and the ease of spoken communication. Speech-language pathologists manually identify articulation error patterns based on their clinical experience, which is a time-consuming and expensive process. This study proposes an automatic pronunciation error identification system that uses a novel dependence network (DN) approach. In order to derive a subject's articulatory information, a photo naming task is performed to obtain the subject's speech patterns. Based on clinical knowledge about speech evaluation, a DN scheme was used to model the relationships of a test word, a subject, a speech pattern, and an articulation error pattern. To integrate DN into automatic speech recognition (ASR), a pronunciation confusion network is proposed to model the probability of DN and is then used to guide the search space of the ASR. Further, to increase the accuracy of the ASR, an appropriate threshold based on a histogram of pronunciation errors is selected in order to disregard rare pronunciation errors. Finally, the articulation error patterns were well identified by integrating the likelihoods of the DNs of each phoneme. The results of this study indicate that it is feasible to clinically implement this dynamic network approach to achieve satisfactory performance in articulation evaluation.     

基于Android平台的智能英语发音训练系统的设计与研究

随着计算机科学技术的发展,英语学习软件的研发和应用数量也逐渐增多.在英语的学习中,智能英语发音训练是练习英语口语的重要部分,目前在英语的发音训练研究中语音识别技术受到高度的关注.随着移动互联网技术的发展,基于Android平台的便携移动设备作为安装英语发音辅助学习系统的主要工具得到了广泛的应用.本文对Android应用程序和英语教学中的英语发音训练进行了分析和研究,在Android平台的基础之上提出了智能英语发音训练系统设计的方案.     

Temporal stabilization of Video Object Segmentation for 3D-TV applications

Our aim is to insert depth information into an existing 2D video sequence to provide content for 3D-TV applications, which we try to achieve through segmentation of the objects in the given 2D video sequence. To this effect, we present a method for temporal stabilization of video object segmentation algorithms for 3D-TV applications. First, two quantitative measures to evaluate temporal stability without ground-truth are discussed. Then, a pseudo-3D curve evolution method, which spatio-temporally stabilizes the estimated segmentation of a video object is introduced. Temporal stability is achieved by re-distributing existing object segmentation errors such that they will be less disturbing when the scene is rendered and viewed in 3D. Our starting point is the hypothesis that if making segmentation errors is inevitable, these errors should be made in a temporally consistent way for 3D-TV applications. This hypothesis is supported by the experiments, which show that there is significant improvement in segmentation quality both in terms of the objective quantitative measures and in terms of the viewing comfort in subjective perceptual tests. Therefore, it is possible to increase the perceptual object segmentation quality without increasing the actual segmentation accuracy.     

单片机语音合成技术

本文介绍一种基于对汉语语汇中一个专用的、数量极为有限的语音子集进行采集与生成的技术，它具有实用性强、成本低、易于实现等特点。     

A new accurate and precise 3-D segmentation method for skeletal structures in volumetric CT data

We developed a highly automated three-dimensionally based method for the segmentation of bone in volumetric computed tomography (CT) datasets. The multistep approach starts with three-dimensional (3-D) region-growing using local adaptive thresholds followed by procedures to correct for remaining boundary discontinuities and a subsequent anatomically oriented boundary adjustment using local values of cortical bone density. We describe the details of our approach and show applications in the proximal femur, the knee, and the skull. The accuracy of the determination of geometrical parameters was analyzed using CT scans of the semi-anthropomorphic European spine phantom. Depending on the settings of the segmentation parameters cortical thickness could be determined with an accuracy corresponding to the side length of 1 to 2.5 voxels. The impact of noise on the segmentation was investigated by artificially adding noise to the CT data. An increase in noise by factors of two and five changed cortical thickness corresponding to the side length of one voxel. Intraoperator and interoperator precision was analyzed by repeated analysis of nine pelvic CT scans. Precision errors were smaller than 1% for trabecular and total volumes and smaller than 2% for cortical thickness. Intraoperator and interoperator precision errors were not significantly different. Our segmentation approach shows: 1) high accuracy and precision and is 2) robust to noise, 3) insensitive to user-defined thresholds, 4) highly automated and fast, and 5) easy to initialize.     

基于改进DeepLabv3+网络的沥青道路裂缝检测方法

针对传统的语义分割技术对于沥青道路裂缝的检测存在检测精度低、误差大的问题,提出了一个基于改进DeepLabv3+网络的语义分割方法。该方法在编码器阶段,采用轻量级MobileNetv2取代DeepLabv3+的主干网络Xception,从而减少参数量;在解码器阶段,引入双注意力机制以进一步提高网络的分割精度;使用Dice Loss函数与原始交叉熵损失函数混合,以缓解样本中前景和背景不平衡问题。最后以道路实时检测的数据为对象进行了大量的实验,结果表明,该方法与原始DeepLabv3+相比,平均交并比(mIoU)、平均像素精度(mPA)分别提升了8.98%和17.39%。与其他主流语义分割模型相比,改进后的DeepLabv3+在沥青道路裂缝的检测上也取得了较好的效果。     

Sequential cardiac segmentation by seed contour tracking

A sequential cardiac segmentation method based on an active contour model, initial seed contour tracking (SCT) scheme, and phase contrast magnetic resonance imaging (MRI) has been developed to improve the accuracy of automatic sequential segmentation of the myocardial boundaries, especially the endocardial boundary. The performance of the proposed method was assessed by experiments performed on phase contrast MRI data sets from three normal human volunteers. Experimental results showed that the propagation of errors caused by improper positioning of initial seed contours in sequential cardiac segmentation was reduced significantly by the use of the SCT scheme.