基于模糊聚类的语音识别

指出模糊聚类方法影响语间识别的正确率 ,提出用基于分解构造的模糊聚类方法设立语音模式的参考向量 ,说明该方法能提高语章识别的正确率。     

Statistical methods in multi-speaker automatic speech recognition

Automatic speech recognition and understanding (ASR) plays an important role in the framework of man-machine communication. Substantial industrial developments are at present in progress in this area. However, after 40 years or so of efforts several fundamental questions remain open. This paper is concerned with a comparative study of four different methods for multi-speaker word recognition: (i) clustering of acoustic templates, (ii) comparison with a finite state automaton, (iii) dynamic programming and vector quantization, (iv) stochastic Markov sources. In order to make things comparable, the four methods were tested with the same material made up of the ten digits (0 to 9) pronounced four times by 60 different speakers (30 males and 30 females). We will distinguish in our experiments between multi-speaker systems (capable of recognizing words pronounced by speakers that have been used during the training phase of the system) and speaker-independent systems (capable of recognizing words pronounced by speakers totally unknown to the system). Half of the corpus (15 male and 15 female) were used for training, and the remaining part for test.     

工件的释放时间和加工时间具有一致性的单机在线排序问题研究

工件的释放时间和加工时间具有一致性, 是指释放时间大的工件其加工时间不小于释放时间小的工件的加工时间, 即若$r_{i}\geq r_{j}$, 则$p_{i}\geq p_{j}$。本文在该一致性约束下, 研究最小化最大加权完工时间单机在线排序问题, 和最小化总加权完工时间单机在线排序问题, 并分别设计出$\frac{\sqrt{5}+1}{2}$-竞争的最好可能在线算法。     

GA-based learning for a model-based object recognition system

This paper proposes a genetic-algorithm-based learning strategy that models membership functions of the fuzzy attributes of surfaces in a model based machine vision system. The objective function aims at enhancing recognition performance in terms of maximizing the degree of discrimination among classes. As a result, the accuracy of recognizing known instances of objects and generalization capability by recognizing unknown instances of known objects are greatly improved. Performance enhancement is achieved by incorporating an off-line learning mechanism using genetic algorithm in the feedback path of the recognition system.     

隐马尔可夫模型在语音识别中的应用

隐马尔可夫模型 (HMM)的基本技术是语音识别中较为成功的算法 .主要是它具有较强的对时间序列结构的建模能力 .本文首先深入浅出地介绍了 HMM的基本技术和一个基于 HMM的孤立词语音识别系统的构成方法 ,其次 ,基于 HMM尚存有一些缺陷 ,造成语音识别能力较弱 ,为此本文又进一步阐述了语音识别应用中的几种改进的 HMM系统及目前的热点方法—— HMM与 ANN构成的混合网络     

Wavelet-based cepstrum calculation

In this paper we present a new wavelet-based algorithm for low-cost computation of the cepstrum. It can be used for real time precise pitch determination in automatic speech and speaker recognition systems. Many wavelet families are examined to determine the one that works best. The results confirm the efficacy and accuracy of the proposed technique for pitch extraction.     

Minimum classification error training in example based speech and pattern recognition using sparse weight matrices

The Minimum Classification Error (MCE) criterion is a well-known criterion in pattern classification systems. The aim of MCE training is to minimize the resulting classification error when trying to classify a new data set. Usually, these classification systems use some form of statistical model to describe the data. These systems usually do not work very well when this underlying model is incorrect. Speech recognition systems traditionally use Hidden Markov Models (HMM) with Gaussian (or Gaussian mixture) probability density functions as their basic model. It is well known that these models make some assumptions that are not correct. In example based approaches, these statistical models are absent and are replaced by the pure data. The absence of statistical models has created the need for parameters to model the data space accurately. For this work, we use the MCE criterion to create a system that is able to work together with this example based approach. Moreover, we extend the locally scaled distance measure with sparse, block diagonal weight matrices resulting in a better model for the data space and avoiding the computational load caused by using full matrices. We illustrate the approach with some example experiments on databases from pattern recognition and with speech recognition.     

层次分析法在身份鉴别方案选择中的应用

针对身份鉴别方案比较的问题,采用层次分析法AHP(Analytic HierarchyProcess)建立了身份鉴别方案比较的数学模型,定量分析了指纹识别、人脸识别、语音识别、虹膜识别、静脉红外识别和DNA识别六种生物识别技术在身份鉴别中的综合表现,并做出评价.评价结果对重要场合安全检查中人员身份鉴别方法的选择有一定的指导意义.     

Development of modular neural networks with fuzzy logic response integration for signature recognition

This paper describes a modular neural network (MNN) for the problem of signature recognition. Currently, biometric identification has gained a great deal of research interest within the pattern recognition community. For instance, many attempts have been made in order to automate the process of identifying a person’s handwritten signature, however this problem has proven to be a very difficult task. In this work, we propose an MNN that has three separate modules, each using different image features as input, these are: edges, wavelet coefficients, and the Hough transform matrix. Then, the outputs from each of these modules are combined by using a Sugeno fuzzy integral. The experimental results obtained by using a database of 30 individual’s shows that the modular architecture can achieve a very high 98% recognition accuracy with a test set of 150 images. Therefore, we conclude that the proposed architecture provides a suitable platform to build a signature recognition system.     

Test recognition

We describe the logic approach to pattern recognition; its key notion is a test. Analyzing the tests allows us to construct functionals characterizing the pattern, as well as procedures to compute them. We present qualitative and quantitative properties of tests, functionals, and recognition procedures. Solutions of a series of known problems are also given.     

保障用户认证安全的判别最大熵语音识别机制

为了提高电子商务中用户认证的安全性,提出了一种高识别率的判别最大熵语音识别机制DME.该语音识别方法同时考虑语音与语言两方面的因素,并将语音和语言特征进行有效地结合,在统一的最大熵模型下,实现判别训练,确保观察样本能正确地分配到其对应状态,以提高所训练出的语音模型的正确识别率.详细的实验及与现有方法的比较结果表明,对不同环境下的语音数据,提出的语音识别方法具有更好的识别性能,对提高电子商务中用户认证系统的安全性能具有理论与实际意义.     

Pattern recognition based Kalman filter for indoor localization using TDOA algorithm

In this paper, we will present a motion pattern recognition based Kalman filter (PRKF), and apply it to the time difference of arrival (TDOA) algorithm of indoor localization. The state matrix in Kalman filter (KF) is determined by the motion pattern which the target node is supposed to act, and this will bring new system error if the assumption is not correct. Considering this, we first create three fuzzy sets using three KFs whose state matrix stand for different motion patterns, then linearly combined the memberships of a target node of the fuzzy sets. Finally, simulation results show that the PRKF can enhance the localization accuracy about more than 20%.     

ABC-PSO算法优化混合核SVM参数及应用

针对混合核支持向量机(SVM)中的可调参数一般是根据经验或人工随机调试得到,不能确保参数最优的局限性,提出用粒子群和人工蜂群的并行混合优化(ABC-PSO)算法来优化混合核SVM参数,找出满足条件的最优参数组合.将该SVM模型应用到语音识别中,通过对三个不同语种的语音数据库的实验仿真,验证了混合算法优化SVM参数所得的优化SVM模型比PSO算法优化SVM所得的模型,具有良好的泛化能力和语音识别能力.     

A generalized preimage for the digital analytical hyperplane recognition

A new digital hyperplane recognition method is presented. This algorithm allows the recognition of digital analytical hyperplanes, such as Naive, Standard and Supercover ones. The principle is to incrementally compute in a dual space the generalized preimage of the ball set corresponding to a given hypervoxel set according to the chosen digitization model. Each point in this preimage corresponds to a Euclidean hyperplane the digitization of which contains all given hypervoxels. An advantage of the generalized preimage is that it does not depend on the hypervoxel locations. Moreover, the proposed recognition algorithm does not require the hypervoxels to be connected or ordered in any way.     

Supervised fuzzy pattern recognition

This paper is devoted to the problem of supervised fuzzy pattern recognition. The cases with non-fuzzy and fuzzy labels are considered. Based on the properties of linearly separable fuzzy classes, some algorithms are proposed for building matching functions of these classes. All algorithms are computer oriented and can be implemented for the automatic recognition of fuzzy patterns.     

Correctness of the linear closure of hyperplane-based recognition algorithms

A theorem concerning the correctness of the linear closure of the set of recognition algorithms based on the separation of objects by hyperplanes is proved.     

On the correctness of a linear closure of set of recognition algorithms of the decision tree type

A recognition operator of the decision tree type is proposed in this paper that is used for the canonical representation of a recognition algorithm, and a theorem is proved about the correctness of the linear closure of a class of incorrect recognition algorithms based on the construction of decision trees.Translated from Dinamicheskie Sistemy, No. 5, pp. 91–94, 1986.     

An analysis of convergence for a learning version of the subspace method

The learning subspace method of pattern recognition has been earlier introduced by Kohonen et al. in a speech recognition application, where the phonemes to be classified are given as spectral representations. In that method, the class subspaces are updated recursively using special rotation matrices, which depend on the training vectors entering one at a time. Here the learning algorithm based on these operators is represented in a general mathematical form, and almost sure convergence is shown to a given criterion that is a function of the statistics of the training set as well as of a set of nonrandom but free parameters. The proof employs current techniques in stochastic approximation theory. For illustration, the resulting classification criterion is then applied to a concrete pattern recognition situation with suitably chosen parameter values.     

Universality of deep convolutional neural networks

Deep learning has been widely applied and brought breakthroughs in speech recognition, computer vision, and many other domains. Deep neural network architectures and computational issues have been well studied in machine learning. But there lacks a theoretical foundation for understanding the approximation or generalization ability of deep learning methods generated by the network architectures such as deep convolutional neural networks. Here we show that a deep convolutional neural network (CNN) is universal, meaning that it can be used to approximate any continuous function to an arbitrary accuracy when the depth of the neural network is large enough. This answers an open question in learning theory. Our quantitative estimate, given tightly in terms of the number of free parameters to be computed, verifies the efficiency of deep CNNs in dealing with large dimensional data. Our study also demonstrates the role of convolutions in deep CNNs.