首页 | 本学科首页   官方微博 | 高级检索  
相似文献
 共查询到20条相似文献,搜索用时 93 毫秒
1.
采用归一化补偿变换的与文本无关的说话人识别   总被引:10,自引:0,他引:10  
在噪声环境下,特别是当说话人识别最常用的模型——高斯混合模型(GMM)失配的情况下,需要对其输出帧似然概率的统计特性进行补偿。文章根据说话人识别的声学特性,提出了一种非线性变换方法——归一化补偿变换。理论分析和实验结果表明:与常用的最大似然(ML)变换相比,该变换能够提高系统识别率,最大可达3.7%,同时可降低误识率,最大可达45.1%。结果说明归一化补偿变换方法基本克服了在与文本无关说话人识别系统中,当说话人的个性特征不断变化、语音与噪声不能很好地分离或者降噪算法对语音有损伤、模型不能很好地匹配时,需要对模型输出的似然概率(得分)进行补偿的局限。这也说明对模型输出的似然概率进行处理是降低噪声和干扰的影响、提高说话人识别率的有效方法。  相似文献   

2.
基于回归分析的语音识别快速自适应算法   总被引:2,自引:2,他引:0  
从回归分析的角度推导出最大似然线性回归算法的等价算法--最小二乘线性回归算法,以及相应的多元线性回归模型。该模型中回归因子间存在着多重共线性,它导致了算法在自适应数据很少时失效。为减轻多重共线性的影响,提出改进算法:伪自适应数据算法。实验表明,当仅有1s-3s自适应数据时,新算法使得系统误识率相对下降2%-6%,随着自适应数据增多,其性能与最大似然线性回归(或最小二乘线性同归)算法趋于一致。  相似文献   

3.
由于训练环境和识别环境的失配,识别系统的性能会严重下降。为此,提出了基于最大后验概率非线性变换的环境自适应算法,可以减小由于环境的失配所引起的系统性能的下降。在本算法中,利用分段线性回归近似非线性变换将训练环境下隐马尔可夫模型(HMM)的均值向量变换到识别环境,减小环境的失配,变换参数的估计采用了最大后验概率估计(MAP)。数字语音识别实验证明:该环境自适应算法的识别性能优于MLST,MAPLR和MLLR等算法。  相似文献   

4.
提出了一种文本无关说话人识别的全特征矢量集模型及互信息评估方法,该模型通过对一组说话人语音数据在特征空间进行聚类而形成,全面地反映了说话人语音的个性特征。对于说话人语音的似然度计算与判决,则提出了一种互信息评估方法,该算法综合分析距离空间和信息空间的似然度,并运用最大互信息判决准则进行识别判决。实验分析了线性预测倒谱系数(LPCC)和Mel频率倒谱系数(MFCC)两种情况下应用全特征矢量集模型和互信息评估算法的说话人识别性能,并与高斯混合模型进行了比较。结果表明:全特征矢量集模型和互信息评估算法能够充分反映说话人语音特征,并能够有效评估说话人语音特征相似程度,具有很好的识别性能,是有效的。  相似文献   

5.
基于分段模糊c-均值的连续密度HMM语音识别模型参数估计   总被引:10,自引:0,他引:10  
本文在分析了连续密度隐马尔可夫模型(CDHMM)的两种主要最大似然参数估计方法的基础上,引入模糊集思想,提出了分段模糊c-均值算法代替普通的分段k-均值算法进行CDHMM的最大似然参数估计。文中给出了其实现方法。实验结果证明其在语音识别中具有很好的性能。  相似文献   

6.
快速准确地识别污染气体种类是光谱法环境监测技术对分类器的基本要求。分段线性分类器简单、计算量小,可以较好的逼近非线性分界面。文章根据最大化分类间隔的思想,结合分段线性分类器和线性支持向量机,设计了单边分段线性分类器优化算法。对某气体模拟剂光谱的分类实验表明,经过优化算法训练的分段线性分类器可以用较少的超平面逼近非线性分界面,而且得到更高的识别准确率。  相似文献   

7.
将矢量泰勒级数(Vector Taylor Series,VTS)特征补偿算法应用于说话人识别,给出了卷积噪声方差的近似闭式解,构建了联合快速估计卷积噪声和加性噪声均值和方差的框架。该算法可在无需失配环境先验信息的前提下,直接从失配语音中估计出卷积噪声和加性噪声的均值和方差,实现对环境失配的补偿。实验结果表明,在信道变化较大的无线信道下,卷积噪声方差的补偿最高可降低误识率3.24%.提升了系统的识别性能。在存在加性噪声的无线信道下,与基于线性失真模型的特征映射算法和倒谱均值减算法相比,本文算法可分别最大降低49.65%和68.06%的误识率,适合于信道变化较大的失配环境补偿。  相似文献   

8.
提出了基于极大似然线性回归(MLLR)调整的说话人模型合成和特征映射方法。MAP调整事后确定相应模型间线性关系,变换参数人为确定;而MLLR调整首先确定相应模型间线性关系,变换参数由训练数据确定,并且可以只调整均值向量。模型合成时,MLLR调整指定通用信道背景模型参数间的线性变换;特征映射时,MLLR调整指定Root GMM-UBM与通用信道背景模型参数间的线性变换。通过对模型参数进行分组调整,可以在训练数据和参数数目间达成平衡。实验结果表明,合适选取MLLR回归类,可以取得比相应MAP调整方法更好的识别效果。  相似文献   

9.
基于Bernstein多项式的自适应混沌时间序列预测算法   总被引:3,自引:0,他引:3       下载免费PDF全文
闫华  魏平  肖先赐 《物理学报》2007,56(9):5111-5118
提出了利用Bernstein多项式对混沌时间序列的动力学方程进行建模的方法,并将该方法与递推最小二乘(RLS)算法相结合,从而可以自适应地逼近混沌时间序列的动力学特性,以达到预测的目的.理论分析和仿真实验表明该方法对一些常见的混沌时间序列具有较高的预测精度和较理想的准确预测率.由于RLS算法的收敛速度较快,因此该方法比较适合于对短混沌时间序列进行实时预测. 关键词: 混沌 预测 Bernstein多项式 RLS算法  相似文献   

10.
陈华  李陶深  赵进创 《光学学报》2007,27(3):424-429
为了对图像复原算法频谱恢复特性进行分析和评价,提出了一种基于高斯函数假设的分析新方法。该方法假设光学传递函数H和退化图像频谱函数G为高斯函数,采用方差以及提出的方差比作为频谱宽度指标,对图像复原算法的频谱恢复特性进行定量分析和评价。分析中对H和G曲线设定两组方差,分无噪声和有噪声两种情况,计算出约束最小平方滤波法(CLS)和最大似然法(PML)等图像复原算法复原的图像频谱曲线及其方差和方差比,采用计算结果对复原算法进行定量的分析和评价,获得良好的效果。分析指出,最大似然法的频谱外推能力和噪声抑制特性均明显好于约束最小平方滤波法。对两种算法的分析评价实验表明,高斯函数假设分析方法是一种简便有效的图像频谱恢复特性分析方法。  相似文献   

11.
The linear hypothesis is the main disadvantage of maximum likelihood linear regression (MLLR).This paper applies the polynomial regression method to model adaptation and establishes a nonlinear model adaptation algorithm using maximum likelihood polynomial regression(MLPR)for robust speech recognition.In this algorithm,the nonlinear relationship between training and testing Gaussian means in every Mel channel is approximated by a set of polynomials and the polynomial coefficients are estimated from adaptation data in test environment using the expectation-maximization(EM)algorithm and maximum likelihood(ML) criterion.The experimental results show that the second-order polynomial can approximate the actual nonlinear function better and in noise compensation and speaker adaptation,the word error rates of MLPR are significantly lower than those of MLLR.The proposed MLPR algorithm overcomes the limitation of linear hypothesis well and can decrease the impact of noise,speaker and other factors simultaneously.It is especially suitable for joint adaptation of speaker and noise.  相似文献   

12.
In this paper, we propose a model adaptation algorithm based on maximum likelihood subband polynomial regression (MLSPR) for robust speech recognition. In this algorithm, the cepstral mean vectors of prior trained hidden Markov models (HMMs) are converted to the log-spectral domain by the inverse discrete cosine transform (DCT) and each log-spectral mean vector is divided into several subband vectors. The relationship between the training and testing subband vectors is approximated by a polynomial function. The polynomial coefficients are estimated from adaptation data using the expectation–maximization (EM) algorithm under the maximum likelihood (ML) criterion. The experimental results show that the proposed MLSPR algorithm is superior to both the maximum likelihood linear regression (MLLR) adaptation and maximum likelihood subband weighting (MLSW) approach. In the MLSPR adaptation, only a very small amount of adaptation data is required and therefore it is more useful for fast model adaptation.  相似文献   

13.
Aiming at the linear temperature compensation algorithm’s disadvantage of temperature measurement error in rotary kiln infrared scanning temperature measurement process, this paper proposes a precise nonlinear cubic polynomial fitting temperature compensation algorithm. The proposed algorithm compensates the temperature values of scanning points on rotary kiln surface by following steps: Calculating temperature difference between the real temperature value of rotary kiln and temperature value measured by infrared scanning temperature measurement system; Fitting the temperature difference data with cubic polynomial; Using the obtained function to compensate temperature. Experimental result shows that compared with the usual linear temperature compensation algorithm, the accuracy of proposed algorithm has raised about 2.25 times when cubic polynomial is used.  相似文献   

14.
One-dimensional analysis of electron acceleration by a nonlinear chirped Gaussian laser pulse was investigated numerically. Two main nonlinear chirped pulses, polynomial and periodical were used. The maximum energy of electron in nonlinear chirped is approximately three times more than that of linear chirped. In the case of first order nonlinear polynomial chirp, Ω(ξ) = 1 + + 2, the electron can be accelerated up to 5.3 GeV. Indeed, the analysis of the electron trajectory in x-z plane showed that the electron in the field of the nonlinear chirped pulse has a much smaller divergence than that of linear chirped pulse.  相似文献   

15.
This paper investigates the problem of modal parameter estimation of time-varying structures under unknown excitation. A time–frequency-domain maximum likelihood estimator of modal parameters for linear time-varying structures is presented by adapting the frequency-domain maximum likelihood estimator to the time–frequency domain. The proposed estimator is parametric, that is, the linear time-varying structures are represented by a time-dependent common-denominator model. To adapt the existing frequency-domain estimator for time-invariant structures to the time–frequency methods for time-varying cases, an orthogonal polynomial and z-domain mapping hybrid basis function is presented, which has the advantageous numerical condition and with which it is convenient to calculate the modal parameters. A series of numerical examples have evaluated and illustrated the performance of the proposed maximum likelihood estimator, and a group of laboratory experiments has further validated the proposed estimator.  相似文献   

16.
17.
A new reliable algorithm based on an adaptation of the standard Homotopy-Perturbation Method (HPM) is presented. The HPM is treated, for the first time, as an algorithm in a sequence of intervals (i.e., time step) for finding accurate approximate solutions of linear and nonlinear systems of ODEs. Numerical comparisons between the Multistage Homotopy-Perturbation Method (MHPM) and the available exact solution and the classical fourth-order Runge-Kutta (RK4) method reveal that the new technique is a promising tool for linear and nonlinear systems of ODEs.  相似文献   

18.
Detecting defects in thin film transistor liquid crystal display (TFT-LCD) panels can be difficult due to non-uniform background brightness and slightly different brightness levels between the defective regions and the background. One well-known method is to inspect the defects of TFT-LCD panels using the polynomial approximation method. However, there are some problems such as determination of the polynomial degree and ghost objects. In this paper, to overcome these problems, we propose a new algorithm using the piecewise linear regression twice and Niblack’s method which do not require to determine the polynomial degree for the approximation. Our algorithm effectively removed ghost objects. The experimental results show that our proposed method performed well and was good enough to substitute the polynomial approximation method.  相似文献   

19.
The traditional linear regression model that assumes normal residuals is applied extensively in engineering and science. However, the normality assumption of the model residuals is often ineffective. This drawback can be overcome by using a generalized normal regression model that assumes a non-normal response. In this paper, we propose regression models based on generalizations of the normal distribution. The proposed regression models can be used effectively in modeling data with a highly skewed response. Furthermore, we study in some details the structural properties of the proposed generalizations of the normal distribution. The maximum likelihood method is used for estimating the parameters of the proposed method. The performance of the maximum likelihood estimators in estimating the distributional parameters is assessed through a small simulation study. Applications to two real datasets are given to illustrate the flexibility and the usefulness of the proposed distributions and their regression models.  相似文献   

20.
Origin是一款制图和数据处理功能强大的应用软件。本文结合大学物理实验,以实例详细介绍如何利用Origin进行实验数据的拟合分析。包括线性回归拟合、多项式拟合和非线性拟合中的一阶指数衰减拟合。实践证明,利用Origin软件对实验数据进行曲线拟合和分析能够显著提高数据处理效率。  相似文献   

设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号