提出了一种适用于抗噪声语音识别的方法,其特征提取过程基于最小方差无失真响应(M in im um varianced istortion less response,M VDR)谱估计方法,并对该特征进行频率弯折以提高其知觉分辨率,最后使用基于正则相关分析的谱变换补偿(C a...提出了一种适用于抗噪声语音识别的方法,其特征提取过程基于最小方差无失真响应(M in im um varianced istortion less response,M VDR)谱估计方法,并对该特征进行频率弯折以提高其知觉分辨率,最后使用基于正则相关分析的谱变换补偿(C anon ica l corre lation based on com pensation,CCBC)法对该特征进行自适应处理,从而提高了系统的鲁棒性。在展览馆噪声、人群噪声和汽车噪声下,与基于传统M e l倒谱系数(M FCC)特征的系统进行了对比实验,结果表明使用本文方法的语音识别系统的识别率得到了显著的提高。展开更多
This paper considers the upper orthant and extremal tail dependence indices for multivariate t-copula. Where, the multivariate t-copula is defined under a correlation structure. The explicit representations of the tai...This paper considers the upper orthant and extremal tail dependence indices for multivariate t-copula. Where, the multivariate t-copula is defined under a correlation structure. The explicit representations of the tail dependence parameters are deduced since the copula of continuous variables is invariant under strictly increasing transformation about the random variables, which are more simple than those obtained in previous research. Then, the local monotonicity of these indices about the correlation coefficient is discussed, and it is concluded that the upper extremal dependence index increases with the correlation coefficient, but the monotonicity of the upper orthant tail dependence index is complex. Some simulations are performed by the Monte Carlo method to verify the obtained results, which are found to be satisfactory. Meanwhile, it is concluded that the obtained conclusions can be extended to any distribution family in which the generating random variable has a regularly varying distribution.展开更多
In standard canonical correlation analysis (CCA), the data from definite datasets are used to estimate their canonical correlation. In real applications, for example in bilingual text retrieval, it may have a great po...In standard canonical correlation analysis (CCA), the data from definite datasets are used to estimate their canonical correlation. In real applications, for example in bilingual text retrieval, it may have a great portion of data that we do not know which set it belongs to. This part of data is called unlabeled data, while the rest from definite datasets is called labeled data. We propose a novel method called regularized canonical correlation analysis (RCCA), which makes use of both labeled and unlabeled samples. Specifically, we learn to approximate canonical correlation as if all data were labeled. Then, we describe a generalization of RCCA for the multi-set situation. Experiments on four real world datasets, Yeast, Cloud, Iris, and Haberman, demonstrate that, by incorporating the unlabeled data points, the accuracy of correlation coefficients can be improved by over 30%.展开更多
文摘提出了一种适用于抗噪声语音识别的方法,其特征提取过程基于最小方差无失真响应(M in im um varianced istortion less response,M VDR)谱估计方法,并对该特征进行频率弯折以提高其知觉分辨率,最后使用基于正则相关分析的谱变换补偿(C anon ica l corre lation based on com pensation,CCBC)法对该特征进行自适应处理,从而提高了系统的鲁棒性。在展览馆噪声、人群噪声和汽车噪声下,与基于传统M e l倒谱系数(M FCC)特征的系统进行了对比实验,结果表明使用本文方法的语音识别系统的识别率得到了显著的提高。
基金The National Natural Science Foundation of China(No.11001052,11171065)the National Science Foundation of Jiangsu Province(No.BK2011058)the Science Foundation of Nanjing University of Posts and Telecommunications(No.JG00710JX57)
文摘This paper considers the upper orthant and extremal tail dependence indices for multivariate t-copula. Where, the multivariate t-copula is defined under a correlation structure. The explicit representations of the tail dependence parameters are deduced since the copula of continuous variables is invariant under strictly increasing transformation about the random variables, which are more simple than those obtained in previous research. Then, the local monotonicity of these indices about the correlation coefficient is discussed, and it is concluded that the upper extremal dependence index increases with the correlation coefficient, but the monotonicity of the upper orthant tail dependence index is complex. Some simulations are performed by the Monte Carlo method to verify the obtained results, which are found to be satisfactory. Meanwhile, it is concluded that the obtained conclusions can be extended to any distribution family in which the generating random variable has a regularly varying distribution.
基金Project (No. 5959438) supported by Microsoft (China) Co., Ltd
文摘In standard canonical correlation analysis (CCA), the data from definite datasets are used to estimate their canonical correlation. In real applications, for example in bilingual text retrieval, it may have a great portion of data that we do not know which set it belongs to. This part of data is called unlabeled data, while the rest from definite datasets is called labeled data. We propose a novel method called regularized canonical correlation analysis (RCCA), which makes use of both labeled and unlabeled samples. Specifically, we learn to approximate canonical correlation as if all data were labeled. Then, we describe a generalization of RCCA for the multi-set situation. Experiments on four real world datasets, Yeast, Cloud, Iris, and Haberman, demonstrate that, by incorporating the unlabeled data points, the accuracy of correlation coefficients can be improved by over 30%.
