基于HMM和LVQ网络混合模型的语音识别方法

提出一种基于隐马尔可夫模型(HMM)和学习向量量化(LVQ)神经网络的语音识别方法.该方法先用HMM生成最佳语音状态序列,然后用函数逼近技术产生对最佳状态序列进行时闻归正,最后通过LVQ神经网络进行分类识别.理论和实验结果表明,混合模型的识别率明显高于隐马尔可夫模型的识别率.     

基于小波网络和HMM的语音识别方法

利用隐马尔可夫模型(HMM)的动态时间序列建模能力及神经网络的模式分类能力,构成混合语音识别模型,同时考虑到语音信号的非平稳性,采用小波分析方法提取语音特征向量。通过时间规整方法,将所有具有可变长度的语音特征向量转换为相同维数的特征向量,从而简化了神经网络的结构。仿真结果表明,采用混合语音识别模型以及时间规整方法,不仅可提高识别率,同时大大缩减了训练时间,获得了很好的识别效果。     

基于循环神经网络的语音识别研究

语音识别作为人工智能研究中不可或缺的一部分已经逐渐渗透到人们的日常生活中。针对传统语音识别方法不能很好地实现并识别复杂多变、非特定人语音的问题,文中提出利用在时间序列上关联性较强的循环神经网络(RNN)建立语音识别模型。考虑到语音信号丰富的时频信息表达,在特征提取环节进行改进,利用具有较好时频分辨率的小波变换(WT)取代快速傅里叶变换(FFT)作为该模型的输入;然后,采用随时间展开的反向传播算法(BPTT)进行特征学习与训练。在实验测试中,首先,对比分析了基于小波变换的特征提取对识别效果的影响;其次,通过与传统的HMM模型及BP神经网络的识别率做对比,验证RNN神经网络可提高语音识别准确率和稳定性。     

MATLAB下的基于HMM模型的语音识别技术的实现

HMM模型(隐含马尔科夫模型)由于对时间序列结构具有较强的建模能力.而逐步成为主流的语音识别技术.该文首先深入浅出地概述了基于HMM的语音识别技术,然后介绍了三个基本问题,最后在MATLAB下实现了孤立词语音识别系统.     

改进的LVQ网络与DTW相结合的语音识别方法

提出一种基于动态时间规整(DTW)和改进的学习矢量量化(LoPLVQ)的神经网络的语音识别方法.该方法用动态时间规整算法先对语音信号进行时间规整,然后通过改进的学习矢量量化神经网络进行语音的分类识别.实验表明,新系统在大规模语音识别方面不仅能缩短训练时间,而且具有较高的识别率.     

利用SVM的聚类算法在时间序列信号识别中的应用

研究了一维时间序列信号识别的问题.针对基于混合高斯模型的隐马尔科夫(HMM)编码准确率低的问题,提出了一种利用多个支持向量机构造混合支持向量机,从而为隐马尔科夫模型提供更精确的观测值编码和发生矩阵,能有效的提高HMM在语音信号识别或者文字识别中的准确率.本方法可以应用到语音识别,文字识别以及生物信息处理等领域.     

基于HMM和ANN的语音情感识别研究

语音情感识别是语音识别中的重要分支,是和谐人机交互的基础理论。由于单一分类器在语音情感识别中的局限性,本文提出了隐马尔科夫模型（HMM）和人工神经网络（ANN）相结合的方法,对高兴、惊奇、愤怒、悲伤、恐惧、平静六种情感分别设计一个HMM模型,得到每种情感的最佳匹配序列,然后利用ANN作为后验分类器对测试样本进行分类,通...     

基于段长分布的HMM语音识别模型

本文针对齐次HMM语音识别模型在使用段长信息时存在的缺陷,形式化地定义了一种适合语音信号描述的自左向右非齐次隐含马尔科夫模型,证明了这种模型的状态转移概率表示与状态段长表示的等效性,并在此基础上提出了基于段长分布的HMM模型(DDBHMM).非特定人连续语音实验结果表明,仅仅利用状态段长信息的DDBHMM语音识别模型比经典HMM模型的性能有了明显的提高(误识率降低了17.8%),展示了DDBHMM的良好的性能,为语音信号的时长、语速、时间断续性以及语音特征的相关性等重要特征的描述和利用开辟了空间.     

从线性预测HMM到一种新的语音识别的混合模型

线性预测HMM(Linear Prediction HMM,LPHMM)并没有象传统HMM那样引入状态输出独立同分布假设,但实用中识别性能并不佳.通过分析两种HMM的各自优劣,本文提出了一种新的语音识别的混合模型,将语音静态特性(基于传统HMM)和动态特性(基于LPHMM)分别描述又有机结合在一起,更为精确地刻划了真实的语音现象,同时又继承使系统的实现改动很小和较小的计算量.汉语大词汇量非特定人连续语音识别的实验表明,混合模型的识别性能显著好于LPHMM和传统HMM.理论上,本文还给出了LPHMM的一组闭式参数重估公式.     

基于概率神经网络的数字音识别

本文主要利用概率神经网络和动态时间规整技术来实现数字音的识别研究。结论是在利用概率神经网络进行语音识别时可以达到比较高的识别率,此外动态时间规整函数的加入,解决了神经网络的模板规整问题。作为语音识别技术的基础,其中包含了小波的基础理论,语音的预处理,DTW技术,端点检测等基础技术。对于神经网络的加入,更加有利于深入了解神经网络这一新兴技术。     

An integrated hybrid neural network and hidden Markov modelclassifier for sonar signals

We present here an integrated hybrid hidden Markov model and neural network (HMM/NN) classifier that combines the time normalization property of the HMM classifier with the superior discriminative ability of the neural net (NN). In the proposed classifier, a left-to-right HMM module is used first to segment the observation sequence of every exemplar into a fixed number of states. Subsequently, all the frames belonging to the same state are replaced by one average frame. Thus, every exemplar, irrespective of its time-state variation, is transformed into a fixed number of frames, i.e., a static pattern. The multilayer perceptron (MLP) neural net is then used as the classifier for these time-normalized exemplars. Some experimental results using sonar biologic signals are presented to demonstrate the superiority of the hybrid integrated classifier     

Basis-Based Speaker Adaptation Using Partitioned HMM Mean Parameters of Training Speaker Models

This paper presents the basis-based speaker adaptation method that includes approaches using principal component analysis (PCA) and two-dimensional PCA (2DPCA). The proposed method partitions the hidden Markov model (HMM) mean vectors of training models into subvectors of smaller dimension. Consequently, the sample covariance matrix computed using the partitioned HMM mean vectors has various dimensions according to the dimension of the subvectors. From the eigen-decomposition of the sample covariance matrix, basis vectors are constructed. Thus, the dimension of basis vectors varies according to the dimension of the sample covariance matrix, and the proposed method includes PCA and 2DPCA-based approaches. We present the adaptation equation in both the maximum likelihood (ML) and maximum a posteriori (MAP) frameworks. We perform continuous speech recognition experiments using the Wall Street Journal (WSJ) corpus. The results show that the model with basis vectors whose dimensions are between those of PCA and 2DPCA-based approaches shows good overall performance. The proposed approach in the MAP framework shows additional performance improvement over the ML counterpart when the number of adaptation parameters is large but the amount of available adaptation data is small. Furthermore, the performance of the approach in the MAP framework approach is less sensitive to the choice of model order than the ML counterpart.     

Training a Hidden Markov Model with a Bayesian Spiking Neural Network

It is of some interest to understand how statistically based mechanisms for signal processing might be integrated with biologically motivated mechanisms such as neural networks. This paper explores a novel hybrid approach for classifying segments of sequential data, such as individual spoken works. The approach combines a hidden Markov model (HMM) with a spiking neural network (SNN). The HMM, consisting of states and transitions, forms a fixed backbone with nonadaptive transition probabilities. The SNN, however, implements a biologically based Bayesian computation that derives from the spike timing-dependent plasticity (STDP) learning rule. The emission (observation) probabilities of the HMM are represented in the SNN and trained with the STDP rule. A separate SNN, each with the same architecture, is associated with each of the states of the HMM. Because of the STDP training, each SNN implements an expectation maximization algorithm to learn the emission probabilities for one HMM state. The model was studied on synthesized spike-train data and also on spoken word data. Preliminary results suggest its performance compares favorably with other biologically motivated approaches. Because of the model’s uniqueness and initial promise, it warrants further study. It provides some new ideas on how the brain might implement the equivalent of an HMM in a neural circuit.     

The Hidden Markov Model of co-articulation and its application to the continuous speech recognition

The co-articulation is one of the main reasons that makes the speech recognition difficult. However, the traditional Hidden Markov Models(HMM) can not model the co-articulation, because they depend on the first-order assumption. In this paper, for modeling the co-articulation, a more perfect HMM than traditional first order HMM is proposed on the basis of the authors' previous works(1997, 1998) and they give a method in that this HMM is used in continuous speech recognition by means of multilayer perceptrons(MLP), i.e. the hybrid HMM/MLP method with triple MLP structure. The experimental result shows that this new hybrid HMM/MLP method decreases error rate in comparison with authors' previous works.     

Maximum likelihood hidden Markov modeling using a dominant sequenceof states

Approximate maximum likelihood (ML) hidden Markov modeling using the most likely state sequence (MLSS) is examined and compared with the exact ML approach that considers all possible state sequences. It is shown that for any hidden Markov model (HMM), the difference between the approximate and the exact normalized likelihood functions cannot exceed the logarithm of the number of states divided by the dimension of the output vectors (frame length). Furthermore, for Gaussian HMMs and a given observation sequence, the MLSS is typically the sequence of nearest neighbor states in the Itakura-Saito sense, and the posterior probability of any state sequence which departs from the MLSS in a single time instant, decays exponentially with the frame length. Hence, for a sufficiently large frame length the exact and approximate ML approach provide similar model estimates and likelihood values     

A Novel Method for the Diagnosis of the Incipient Faults in Analog Circuits Based on LDA and HMM

Diagnosis of incipient faults for electronic systems, especially for analog circuits, is very important, yet very difficult. The methods reported in the literature are only effective on hard faults, i.e., short-circuit or open-circuit of the components. For a soft fault, the fault can only be diagnosed under the occurrence of large variation of component parameters. In this paper, a novel method based on linear discriminant analysis (LDA) and hidden Markov model (HMM) is proposed for the diagnosis of incipient faults in analog circuits. Numerical simulations show that the proposed method can significantly improve the recognition performance. First, to include more fault information, three kinds of original feature vectors, i.e., voltage, autoregression-moving average (ARMA), and wavelet, are extracted from the analog circuits. Subsequently, LDA is used to reduce the dimensions of the original feature vectors and remove their redundancy, and thus, the processed feature vectors are obtained. The LDA is further used to project three kinds of the processed feature vectors together, to obtain the hybrid feature vectors. Finally, the hybrid feature vectors are used to form the observation sequences, which are sent to HMM to accomplish the diagnosis of the incipient faults. The performance of the proposed method is tested, and it indicates that the method has better recognition capability than the popularly used backpropagation (BP) network.     

Speech recognition algorithm based on neural network and hidden Markov model

This study proposes a hybrid model of speech recognition parallel algorithm based on hidden Markov model (HMM) and artificial neural network (ANN). First, the algorithm uses HMM for time-series modeling of speech signals and calculates the voice to the HMM of the output probability score. Second, with the probability score as input to the neural network, the algorithm gets information for classification and recognition and makes a decision based on the hybrid model. Finally, Matlab software is used to train and test sample data. Simulation results show that using the strong time-series modeling ability of HMM and the classification features of neural network, the proposed algorithm possesses stronger noise immunity than the traditional HMM. Moreover, the hybrid model enhances the individual flaws of the HMM and the neural network and greatly improves the speed and performance of speech recognition.     

有序聚类方法及其在神经网络语音识别中的应用

本文提出了一种新的网络结构,我们称之为有序聚类网络。这种网络能够对语音信号进行特征提取,很好地解决神经网络语音识别中的时间规整问题。有序聚类网络从输入语音信号的特征矢量序列中撮出一组固定数目的特 矢量,然后将这组特征矢量馈入神经网络分类器进行识别。和其他的神经网络语音识别方法相比较,用这种网络进行前端处理,可以缩短后端神经网络分类器的训练和识别时间,简化经分类器的网络产高的识别率。根据该 们建立了     

Fast and memory efficient implementation of the exact PNN

Straightforward implementation of the exact pairwise nearest neighbor (PNN) algorithm takes O(N(3)) time, where N is the number of training vectors. This is rather slow in practical situations. Fortunately, much faster implementation can be obtained with rather simple modifications to the basic algorithm. In this paper, we propose a fast O(tauN(2)) time implementation of the exact PNN, where tau is shown to be significantly smaller than N, We give all necessary data structures and implementation details, and give the time complexity of the algorithm both in the best case and in the worst case. The proposed implementation achieves the results of the exact PNN with the same O(N) memory requirement.