基于相空间重构的语音特征研究

本文通过重构语音信号相空间。研究语音的相似序列重复度及其熵信息,分析比较了语音信号在相空间中的非线性特征。根据清音和浊音在多维相空间中的不同空间分布特性,对语音音素进行了分类。利用语音信号在相空间中的非线性特征可以为语音识别研究提供一个新的方向。     

混沌噪声的神经网络建模应用于参数估计问题中

提出了基于神经网络预测器的参数估计算法，该算法将神经网络拟合非线性函数的能力和功率谱分析技术相结合．文中介绍了相空间重建技术和神经网络的原理，对于神经网络预测模型，给出了所提算法的原理和步骤，针对具体应用问题，用计算机仿真实验验证了该算法提取混沌噪声中信号参数的有效性,给出了实验结果和必要的分析．     

基于ELMAN网络的非线性混沌微弱信号检测

为了快速和实时地从具有强噪声的较低信噪比的原始信号中检测出有用信息,设计了一种混沌相空间重构理论和ELMAN神经网络的信号检测方法;首先,描述了采用混沌相空间重构理论对原始信号进行重构的原理和方法,在获取重构的时间序列的基础上,采用ELMAN网络来近似表示用于检测信号的函数型,然后,设计了ELMAN网络中各层之间连接权值的计算方式,并提出了采用ELMAN网络进行信号检测的具体过程,最后给出了采用混沌相空间重构理论和ELMAN网络的信号检测模型;对Lorenz混沌系统模型进行仿真实验,结果证明了文章方法能有效地对瞬时信号和周期性信息进行检测,在具有高斯白噪声的情况下,仍然具有降噪效果好的优点,是一种用于信号检测的可行性方法。     

基于多嵌入维数的风力发电功率组合预测模型

为了减小混沌系统的重构参数对预测结果的影响,提出了基于多嵌入维数的风力发电功率组合预测模型.分别使用线性加权算法和神经网络算法对单一的基于相空间重构的神经网络模型进行组合,既综合了各嵌入维数下的信息,又将各维数下的预测偏差进行融合,从而有效提高了预测精度.通过对黑龙江富锦风电场的功率时间序列进行验证,证实了该组合模型的有效性,神经网络非线性组合算法的预测误差小于7%.     

利用VLRBP神经网络改善汇率预测

分别使用基于滑动窗口的VLRBP神经网络模型和基于C-C相空间重构的VLRBP神经网络模型及ARIMA-GARCH模型对欧元汇率时间序列建模和预测,通过比较发现基于C-C相空间重构的VLRBP神经网络对于含有大量非线性成分的欧元汇率时间序列的预测比较准确。同时,为了提高基于滑动窗口的VLRBP网络的泛化性能,提出在训练VLRBP神经网络时应用浴盆曲线方法选取隐层神经元个数和滑动窗口尺寸。     

SVM在非线性网络流量预测中的应用研究

网络流量是一种高度自相关、非线性时间序列数据,传统预测方法都是基于线性模型,无法反映网络流量的非线性变化规很,导致预测精度不高.为了提高网络流量的预测精度,在分析网络流量特征的基础上,提出一种基于相空间重构的支持向量机网络流量预测模型.首先利用相空间重构对网络流量原始数据进行重构,捕捉原始数据的多样性,然后将重构的数据...     

基于小波神经网络的推进电机系统状态检测

本文将小波分析理论引入到相空间重构方法之中,仿真结果表明：小波相空间重构方法不但可以精确重构出原系统的混沌特征,还可以重构出原系统的周期特征、倍周期特征和拟周期特征等。在小波相空间重构的分析基础之上,建立小波神经网络来分析无刷推进电机系统的运行状态。     

基于小波变换和神经网络的车内噪声信号重构

为获取较高精度车内噪声主动控制(Active Noise Control, ANC)参考信号,提出了一种基于小波变换和BP神经网络的车内噪声信号重构方法。以在某轿车采集到的噪声信号为基础,用声学传递路径分析(TPA)方法确定影响车内噪声的关键点信号。鉴于噪声源信号对车内信号非线性关系的复杂性,建立BP神经网络的噪声重构模型,并利用小波分解来降低噪声信号的非平稳性。为对比重构效果,建立BP神经网络噪声重构模型。结果表明,本文提出算法的重构值与实测值之间的平均绝对误差比BP神经网络小,并且基于小波变换和BP网络重构模型的平均绝对误差均小于0.01。该方法能够对车内噪声信号进行准确、有效的重构。     

一种基于RBF神经网络的预测器模型及其研究

非线性复杂系统的预测控制是一种高性能的控制方法,其关键在于非线性预测器模型的实现。论文从径向基函数(RBF)神经网络原理分析出发,探讨了一种用于神经网络的预测模型设计方法,并将此方法用于实际非线性系统的预测控制。结果表明,基于RBF的神经网络预测模型可快速准确地完成对非线性动态过程的预测描述,因而可以在非线性系统的预测控制中得到良好的应用。     

分形滤波的网络流量预测

提出了一种结合分形滤波与线性神经网络进行网络流量预测的新方法。通过分形滤波增强网络流量中的长相关结构,使序列更加平滑,根据相空间重构理论利用线性神经网络进行预测操作,并用实际网络流量验证该模型的有效性。     

A nonlinear prediction model for Chinese speech signal based on RBF neural network

A novel method for Chinese speech time series prediction model is proposed. In order to reconstruct the phase space of Chinese speech signal, the delay time and embedding dimension are calculated by C–C method and false nearest neighbor algorithm. The maximum lyapunov exponent and correlation dimension of Chinese speech phoneme are calculated by wolf algorithm and genetic programming algorithm. The numerical results show that there exists nonlinear characteristics in Chinese speech signal. Based on the analysis method of RBF neural network and the nonlinear characteristic parameters such as the delay time and embedding dimension, a nonlinear prediction model is designed. In order to further verify the prediction performance of the designed prediction model, waveform comparison and four evaluation indexes are used. It is shown that compared with the linear prediction model and back propagation neural network nonlinear prediction model, prediction error of the RBF neural network nonlinear prediction model is significantly reduced, and the model has higher prediction accuracy and prediction performance.

     

基于差分过零峰值幅度特征的抗噪语音识别

针对多数语音识别系统在噪音环境下性能急剧下降的问题,提出了一种新的语音识别特征提取方法。该方法是建立在听觉模型的基础上,通过组合语音信号和其差分信号的上升过零率获得频率信息,通过峰值检测和非线性幅度加权来获取强度信息,二者组合在一起,得到输出语音特征,再分别用BP神经网络和HMM进行训练和识别。仿真实现了不同信噪比下不依赖人的50词的语音识别,给出了识别的结果,证明了组合差分信息的过零与峰值幅度特征具有较强的抗噪声性能。     

Speech Versus Touch: A Comparative Study of the Use of Speech and DTMF Keypad for Navigation

This article reports on an experiment that critically tests user preference for an input modality (speech vs. Dual Tone Multiple Frequency[DTMF])in a phone-based message retrieval system. Unlike previous studies that compared these two modalities, the speech system used in this study was a fully functioning natural language system, and participants in this study were working professionals, rather than college students. Results indicate that (a) DTMF was more effective and efficient for linear tasks, whereas speech was better for nonlinear tasks; (b) speech was preferred to DTMF by a majority of users; (c) speech was judged as being more satisfying, more entertaining, and easier to use than DTMF; and (d) user preference for a particular modality was better predicted by user performance in nonlinear tasks rather than linear ones. Possible reasons for users' continuing preference for the speech modality even after experiencing fairly high recognition errors are discussed. Finally, the importance of examining speech user interfaces from other perspectives, in addition to efficiency maximization, is emphasized. The results of this study have theoretical, as well as practical, implications for the design of speech user interfaces and interactive voice response applications.     

非线性系统模糊神经网络变结构自适应控制

在非线性系统的模糊动力学模型基础上,提出一种模糊神经网络变结构自适应控制器;网络的结构根据非线性系统特性动态构成,基于该网络提出非线性预测器,基于梯度法提出了一种网络参数学习算法,并分析了收敛性及其性质。将网络预测器与参数学习算法相结合,构成自适应控制算法,证明了算法的收敛性。仿真结果证实了算法的有效性。     

A state prediction scheme for discrete time nonlinear dynamic systems

A state prediction scheme is proposed for discrete time nonlinear dynamic systems with non-Gaussian disturbance and observation noises. This scheme is based upon quantization, multiple hypothesis testing, and dynamic programming. Dynamic models of the proposed scheme are as general as dynamic models of particle predictors, whereas the nonlinear models of the extended Kalman (EK) predictor are linear with respect to the disturbance and observation noises. The performance of the proposed scheme is compared with both the EK predictor and sampling importance resampling (SIR) particle predictor. Monte Carlo simulations have shown that the performances of the proposed scheme, EK predictor, and SIR particle predictor are all model-dependent, that is, one performs better than the others for a given example. Some examples, for which the proposed scheme performs better than the others do, are also given in the paper.     

基于复合正交神经网络的灰色PID控制

结合传统反馈控制方法和灰色预测控制的预测控制器已在控制系统中获得了成功的应用。由于复合正交神经网络具有学习算法简单、收敛速度快,有逼近线性或非线性函数的优良特性。与灰色预测方法相比,神经网络预测精度高,且误差可控,如果把神经网络作为灰色预测器,建立一种灰色预测控制,那么就会在控制系统中获得良好的控制性能。为此,提出一种结合传统的PID控制和神经网络灰色预测补偿的灰色PID控制器,可对系统进行在线灰色估计和控制,由复合正交神经网络对不确定部分建立的灰色预测模型,可根据系统的参数变化来自动调节预测补偿值,使系统响应具有适应性。仿真结果表明,与传统的PID控制方法相比,该控制器可获得更为优良的动态性能和鲁棒性。     

基于MFCC与分形维数混合参数的语音识别

针对语音气流中具有混沌特征,而分形可以定量地分析混沌现象,故分形可用来分析语音信号。语音波形具有分形特征,将分形用于改善语音识别技术更好地表现语音的特征,避免传统的分段线性处理所产生的局限性。将传统特征参数MFCC与分形特征结合起来,组成混合参数用于语音识别。实验结论显示,基于MFCC与分形维数混合参数的语音识别方法要好于使用单一MFCC参数的语音识别方法。     

随机系统中能容忍连续丢包和测量时延的卡尔曼滤波

通过转换原线性系统到能容忍连续丢包和测量时延的随机参数系统,推导了一个最优线性滤波器.给出一个仿真例子,比较已存在的结果,仿真结果表明被提出的线性滤波器有优越的性能.然而,该滤波器不能应用于非线性系统.从应用角度,为非线性系统提出了一个增强型的滤波器.而且,该增强型的滤波器能成功地应用于不可靠的无线传感器网络场景来跟踪移动目标.这些滤波器只依靠测量值的达到概率,而不需要知道某一时刻测量是否接收.仿真说明了被提出的增强型滤波器不仅能改善实时目标跟踪的鲁棒性,而且比标准的扩展卡尔曼滤波器能够提供更精确的估计.     

Stabilization of nonlinear delay systems using approximate predictors and high-gain observers

We provide a solution to the heretofore open problem of stabilization of systems with arbitrarily long delays at the input and output of a nonlinear system using output feedback only. The solution is global, employs the predictor approach over the period that combines the input and output delays, addresses nonlinear systems with sampled measurements and with control applied using a zero-order hold, and requires that the sampling/holding periods be sufficiently short, though not necessarily constant. Our approach considers a class of globally Lipschitz strict-feedback systems with disturbances and employs an appropriately constructed successive approximation of the predictor map, a high-gain sampled-data observer, and a linear stabilizing feedback for the delay-free system. The obtained results guarantee robustness to perturbations of the sampling schedule and different sampling and holding periods are considered. The approach is specialized to linear systems, where the predictor is available explicitly.     

Choosing the parameters of the NARMA model implemented with the recurrent perceptron for speech prediction

Speech signals have statistically nonstationary properties and cannot be processed properly by means of classical linear parametric models (AR, MA, ARMA). The neural network approach to time series prediction is suitable for learning and recognizing the nonlinear nature of the speech signal. We present a neural implementation of the NARMA model (nonlinear ARMA) and test it on a class of speech signals, spoken by both men and women in different dialects of the English language. The Akaike’s information criterion is proposed for the selection of the parameters of the NARMA model.