基于神经网络集成的手写识别系统

在研究基于隐马尔可夫模型的识别器和基于距离分类器的识别器的识别结果基础上,提出两种基于集成神经网络的手写识别系统:比较神经网络识别系统和全排列神经网络识别系统.实验分析表明,该系统对西文手写体的识别率最高可达到99%,比单独使用原始识别器的识别率提高10个百分点,达到了良好的识别效果.     

基于正交实验设计的语音识别特征参数优化

鉴于语音识别性能与所选用的语音特征参数密切相关,提出一种系统性的实用的特征参数优化方法——基于方差的正交实验设计法。首先进行因素(语音特征参数)和水平的选择,再根据数理统计与正交性原理,从大量的实验点中挑选适量的具有代表性、典型性的点构造正交表进行正交实验,最后通过计算对正交实验结果进行分析,找出最优的特征参数组合。与目前参数的简单组合方案相比较,新方法的误识率下降了5.6%,响应时间减少了181.37ms。实验结果表明,正交实验设计用于语音特征参数优化是有效的,对后续研究具有指导意义。     

基于CHMM语音识别特征参数的选择方法

基于CHMM的语音识别系统识别率高,但却占用系统资源较大,从而限制了其在资源受限的实际应用环境的有效实现.针对上述问题,给出特征参数选择的理论依据,弥补以往研究仅从实验结果分析,缺少理论依据的不足;同时提出根据各特征参数对系统误识率的影响程度来选择特征参数的新方法.该方法能使系统在训练,识别过程中的计算量和存储量明显减小,同时系统误识率不会显著改变.这为资源受限的语音识别系统,提供新的思路和有效的特征参数选择方法.     

一种基于贪婪策略的说话人语音特征优选方法

为了提高说话人识别的性能,提出一种语音特征优选方法,从目前使用效果较好的特征参数中,采用贪婪算法优选出若干维特征用于识别。在TIMIT语音数据库上实验显示,识别率相比传统方法提高了1.6%;对于加入了噪声的语音,识别率提高了6%,识别速度提高了5倍左右。实验结果表明,优选后的特征参数能够去除不良特征对识别系统的干扰,有效提高说话人识别系统的识别率和识别速度。     

一种基于鲁棒特征的模型补偿噪声语音识别方法

针对抗噪声语音特征技术和基于MFCC特征的模型补偿技术在低信噪比时识别率不高的缺点，将抗噪声语音特征和模型补偿结合起来，提出了一种基于单边自相关序列(One—sided autocorrelation，OSA)MFCC特征的模型补偿噪声语音识别方法，以提高语音识别系统在低信噪比时的性能。对0～9十个英文数字和NOISEX92中的白噪声、F16噪声和FACTORY噪声的识别实验结果表明．本文提出的识别方法可以有效地提高OSA—MFCC识别器在噪声环境中的识别率，并且在低信噪比时其性能明显优于经过相同补偿处理的MFCC识别器。     

文本无关的说话人识别系统抗噪方法研究

研究了基于美尔倒谱特征参数及高斯混合模型的文本无关的说话人识别系统,为了提高噪声环境下识别系统的识别率,从两个角度研究改善该系统抗噪性能的方法,即利用语音识别将文本无关的系统转化为文本有关的说话人识别方法和通过选择鲁棒性较强的帧进行说话人识别的方法,分析了以上方法对系统识别性能的改善作用,并通过实验验证上述方法确实可以提高系统在噪声环境下的识别率。     

语音信号特征参数的分析和选取

在与文本有关的说话人识别系统中,既需要识别说话人的身份,又需要识别语音文本的内容。语音信号特征参数的选取对系统来说至关重要。目前,在传统语音识别系统的研究中,主要采用MFCC参数作为特征参数进行识别。笔者对语音信号特征参数进行分析,对不同的语音特征参数组合进行实验。实验结果证明,在该系统中,MFCC参数与基音参数的组合提高了系统的识别率。     

基于识别结果反馈信息的闭环联机字符识别系统

文章提出了一种基于识别结果反馈信息的闭环联机字符识别系统。通常的识别系统只是执行了从输入数据空间到识别结果空间的单向映射,可以看作为一种开环识别系统;而文中提出了一种闭环识别系统:它在完成初次的识别之后,还将识别结果信息反馈给系统中的规则库;规则库依据此反馈信息触发相应规则驱动识别器对输入字符进行再次识别,从而得到最终的识别结果。实验结果表明,在抽取的特征和识别方法相同的条件下,闭环识别系统的识别率将高于开环识别系统的识别率。     

一种加权矢量量化的说话人识别技术

该文研究了在基于矢量量化的说话人识别方法中采用加权的失真测度对识别率的影响。在采用加权欧氏距离失真测度时,利用特征参数的离散程度来确定权值,提出了基于标准差的加权失真测度和基于方差的加权失真测度。实验结果表明,在以MFCC为特征参数的说话人识别系统中,使用这两种算法均可以提高识别率。     

一种改进的特征提取方法在语音识别中的应用

针对梅尔频率倒谱系数(MFCC)参数在噪声环境中语音识别率下降的问题,提出了一种基于耳蜗倒谱系数(CFCC)的改进的特征参数提取方法.提取具有听觉特性的CFCC特征参数;运用改进的线性判别分析(LDA)算法对提取出的特征参数进行线性变换,得到更具有区分性的特征参数和满足隐马尔可夫模型(HMM)需要的对角化协方差矩阵;进行均值方差归一化,得到最终的特征参数.实验结果表明:提出的方法能有效地提高噪声环境中语音识别系统的识别率和鲁棒性.     

Rough set-based heuristic hybrid recognizer and its application in fault diagnosis

Rough set theory (RS) has been a topic of general interest in the field of knowledge discovery and pattern recognition. Machine learning algorithms are known to degrade in performance when faced with many features (sometimes attributes) that are not necessary for rule discovery. Many methods for selecting a subset of features have been proposed. However, only one method cannot handle the complex system with many attributes or features, so a hybrid mechanism is proposed based on rough set integrating artificial neural network (Rough-ANN) for feature selection in pattern recognition. RS-based attributes reduction as the preprocessor can decrease the inputs of the NN and improve the speed of training. So the sensitivity of rough set to noise can be avoided and the system’s robustness is to be improved. A RS-based heuristic algorithm is proposed for feature selection. The approach can select an optimal subset of features quickly and effectively from a large database with a lot of features. Moreover, the validity of the proposed hybrid recognizer and solution is verified by the application of practical experiments and fault diagnosis in industrial process.     

Multiple cameras audio visual speech recognition using active appearance model visual features in car environment

Consideration of visual speech features along with traditional acoustic features have shown decent performance in uncontrolled auditory environment. However, most of the existing audio-visual speech recognition (AVSR) systems have been developed in the laboratory conditions and rarely addressed the visual domain problems. This paper presents an active appearance model (AAM) based multiple-camera AVSR experiment. The shape and appearance information are extracted from jaw and lip region to enhance the performance in vehicle environments. At first, a series of visual speech recognition (VSR) experiments are carried out to study the impact of each camera on multi-stream VSR. Four cameras in car audio-visual corpus is used to perform the experiments. The individual camera stream is fused to have four-stream synchronous hidden Markov model visual speech recognizer. Finally, optimum four-stream VSR is combined with single stream acoustic HMM to build five-stream AVSR. The dual modality AVSR system shows more robustness compared to acoustic speech recognizer across all driving conditions.     

一元化特征建模技术研究

特征建模是实现ＣＡＤ／ＣＡＭ以及并行工程中信息集成的键。一元化征建模技术融特征设计和特征识别两者一体，设计阶段的初始设计特征进入应用领域后，经过特征识别器表成领域专用特征。     

基于特征选择的字符识别

特征选择问题是机器学习和模式识别中的一个重要问题。其本质上是一个多因素优化问题。该文将试验设计与多因素优化问题联系起来,利用正交试验设计的统计特性,从特征集中筛选有效鉴别特征子集。在南京理工大学NUST603HW手写汉字库以及Concordia大学的CENPARMI手写体阿拉伯数字数据库上的试验结果表明,所提出的特征选择方法不仅提高了识别率,而且识别结果十分稳定。     

针对带约束非线性规划问题的遗传算法

研究了针对带约束非线性规划问题的遗传算法(CNP-GA),设计了相应的适应度函数及处理约束的方法,结合了排挤和排序选择以保证群体的多样性,通过邻域搜索和变异算子进行联合演化,并利用多个群体的竞争得到全局解,对非线性限制规划例子的测试证明本算法是有效而可行的。     

Influence of erroneous learning samples on adaptation in on-line handwriting recognition

We have considered problems involved in the self-supervised learning process of an on-line handwriting recognition system. Our system is able to recognize isolated characters by comparing them to prototype characters with a method based on the Dynamic Time Warping algorithm. The recognition system is adapted by adding new prototypes, inactivating confusing or erroneous ones, and reshaping existing prototypes with a method based on the Learning Vector Quantization. We have analyzed the sources of erroneous learning samples and studied the influence of such samples on the performance of the recognizer via simulations. In these simulations, two adaptation strategies combined with four methods for inactivating prototypes were applied. The results of the simulations showed that the adaptation strategies are able to improve the system's recognition rate and the prototype inactivation methods do reduce the harmful effects of erroneous learning samples.     

From members to teams to committee-a robust approach to gestural and multimodal recognition

When building a complex pattern recognizer with high-dimensional input features, a number of selection uncertainties arise. Traditional approaches to resolving these uncertainties typically rely either on the researcher's intuition or performance evaluation on validation data, both of which result in poor generalization and robustness on test data. This paper describes a novel recognition technique called members to teams to committee (MTC), which is designed to reduce modeling uncertainty. In particular, the MTC posterior estimator is based on a coordinated set of divide-and-conquer estimators that derive from a three-tiered architectural structure corresponding to individual members, teams, and the overall committee. Basically, the MTC recognition decision is determined by the whole empirical posterior distribution, rather than a single estimate. This paper describes the application of the MTC technique to handwritten gesture recognition and multimodal system integration and presents a comprehensive analysis of the characteristics and advantages of the MTC approach.