采用压缩感知的改进的语音转换算法

提出了一种基于压缩感知的考虑语音帧间信息的语音转换算法。根据连续多帧语音的线谱对参数所构成的矢量在离散余弦变换域具有稀疏性,利用压缩感知技术对该矢量压缩成短矢量,并将该压缩后的短矢量作为特征参数训练语音转换函数。实验测试结果表明,选择合适的语音帧数时,该算法的性能要比传统的采用加权频率卷绕的转换算法提高3.21%。这说明,充分有效地利用语音帧间的相关信息会使转换语音保持更稳定的帧间声学特性,有利于提高语音转换系统的性能,     

A modified voice conversion algorithm using compressed sensing

A voice conversion algorithm,which makes use of the information between continuous frames of speech by compressed sensing,is proposed in this paper.According to the sparsity property of the concatenated vector of several continuous Linear Spectrum Pairs（LSP）in the discrete cosine transformation domain,this paper utilizes compressed sensing to extract the compressed vector from the concatenated LSPs and uses it as the feature vector to train the conversion function.The results of evaluations demonstrate that the performance of this approach can averagely improve 3.21%with the conventional algorithm based on weighted frequency warping when choosing the appropriate numbers of speech frame.The experimental results also illustrate that the performance of voice conversion system can be improved by taking full advantage of the inter-frame information,because those information can make the converted speech remain the more stable acoustic properties which is inherent in inter-frames.     

面向少量语料的语音转换算法

针对目标说话人可能存在语料不足的情况,本文提出了一种有限语料下的统一张量字典语音转换算法。从语料库中选取N个说话人作为语音张量字典的基础说话人,通过多序列动态时间规整算法使这N个说话人的平行语音段对齐,从而建立由N个二维基础字典构成的张量字典。在语音转换阶段,源、目标说话人语音都可以通过张量字典中各基础字典的线性组合,构造出各自的语音字典,实现了语音转换。实验结果表明,当基础说话人个数达到14时,只需要极少的目标说话人语料,便可获得与传统的基于非负矩阵分解转换算法相当的转换效果,这极大地方便了语音转换系统的应用。     

An algorithm for voice conversion with limited corpus

Under the condition of limited target speaker's corpus, this paper proposed an algorithm for voice conversion using unified tensor dictionary with limited corpus. Firstly,parallel speech of N speakers was selected randomly from the speech corpus to build the base of tensor dictionary. And then, after the operation of multi-series dynamic time warping for those chosen speech, N two-dimension basic dictionaries can be generated which constituted the unified tensor dictionary. During the conversion stage, the two dictionaries of source and target speaker were established by linear combination of the N basic dictionaries using the two speakers' speech. The experimental results showed that when the number of the basic speaker was 14, our algorithm can obtain the compared performance of the traditional NMFbased method with few target speaker corpus, which greatly facilitate the application of voice conversion system.     

线阵CCD图像处理算法研究

研究了线阵CCD在动态测量中的整体图像处理方法 ,包括杂散点剔除、图像平滑、边缘识别等方法。提出的整体图像的样条插值方法使得边缘识别精度比较高 ,实现起来较容易 ,该方法特别对复杂环境下的动态CCD图像处理有效     

A global inversion method for multi-dimensional NMR logging

We describe a general global inversion methodology of multi-dimensional NMR logging for pore fluid typing and quantification in petroleum exploration. Although higher dimensions are theoretically possible, for practical reasons, we limit our discussion of proton density distributions as a function of two (2D) or three (3D) independent variables. The 2D can be diffusion coefficient and T(2) relaxation time (D-T(2)), and the 3D can be diffusion coefficient, T(2), and T(1) relaxation times (D-T(2)-T(1)) of the saturating fluids in rocks. Using the contrast between the diffusion coefficients of fluids (oil and water), the oil and water phases within the rocks can be clearly identified. This 2D or 3D proton density distribution function can be obtained from either two-window or regular type multiple CPMG echo trains encoded with diffusion, T(1), and T(2) relaxation by varying echo spacing and wait time. From this 2D/3D proton density distribution function, not only the saturations of water and oil can be determined, the viscosity of the oil and the gas-oil ratio can also be estimated based on a previously experimentally determined D-T(2) relationship.     

数字式声呐显示系统的GSC算法

方位历程显示已被证明为检测信号的有效手段。数字式声呐的多波束数据与调亮显示系统之间存在着一个接口,它有可能使信号处理系统已获得的增益受到损失。选择合理的算法会使这种损失减到最小。本文提出的灰度级转换算法(GSC)是一种实时的数字运算技术。在输入信噪比较低时会使弱信号的检测能力有所改善,同时又对强信号的检测不会有严重的影响。本文提出的算法易于由硬件实现。计算机模拟实验的结果与理论一致。文中给出了硬件设计的概要说明。     

A high-order multi-dimensional HLL-Riemann solver for non-linear Euler equations

This article presents a numerical model that enables to solve on unstructured triangular meshes and with a high-order of accuracy, a multi-dimensional Riemann problem that appears when solving hyperbolic problems.For this purpose, we use a MUSCL-like procedure in a “cell-vertex” finite-volume framework. In the first part of this procedure, we devise a four-state bi-dimensional HLL solver (HLL-2D). This solver is based upon the Riemann problem generated at the centre of gravity of a triangular cell, from surrounding cell-averages. A new three-wave model makes it possible to solve this problem, approximately. A first-order version of the bi-dimensional Riemann solver is then generated for discretizing the full compressible Euler equations.In the second part of the MUSCL procedure, we develop a polynomial reconstruction that uses all the surrounding numerical data of a given point, to give at best third-order accuracy. The resulting over determined system is solved by using a least-square methodology. To enforce monotonicity conditions into the polynomial interpolation, we develop a simplified central WENO (CWENO) procedure.Numerical tests and comparisons with competing numerical methods enable to identify the salient features of the whole model.     

采用联合字典优化的噪声鲁棒性语音转换算法

针对含噪语音难以实现有效的语音转换,本文提出了一种采用联合字典优化的噪声鲁棒性语音转换算法。在联合字典的构成中,语音字典采用后向剔除算法(Backward Elimination algorithm,BE)进行优化,同时引入噪声字典,使得含噪语音与联合字典相匹配。实验结果表明,在保证转换效果的前提下,后向剔除算法能够减少字典帧数,降低计算量。在低信噪比和多种噪声环境下,本文算法与传统NMF算法和基于谱减法消噪的NMF转换算法相比具有更好的转换效果,噪声字典的引入提升了语音转换系统的噪声鲁棒性。     

基于最佳线性数据融合的双基地声呐定位优化算法

本文提出了基于最佳线性数据融合的双基地声呐定位优化算法。给出了算法的原理，并通过数值仿真，对不同条件下，双基地系统定位误差的几何分布进行了研究。仿真研究结果表明，采用最佳线性数据融合的双基地声呐定位优化算法，可以更好地利用双基地声呐测量的冗余信息，有效改善双基地声呐中基线区域的定位精度。     

A new approach of a limiting process for multi-dimensional flows

An enhanced Multi-dimensional Limiting Process (e-MLP) is developed for the accurate and efficient computation of multi-dimensional flows based on the Multi-dimensional Limiting Process (MLP). The new limiting process includes a distinguishing step and an enhanced multi-dimensional limiting process. First, the distinguishing step, which is independent of high order interpolation and flux evaluation, is newly introduced. It performs a multi-dimensional search of a discontinuity. The entire computational domain is then divided into continuous, linear discontinuous and nonlinear discontinuous regions. Second, limiting functions are appropriately switched according to the type of each region; in a continuous region, there are no limiting processes and only higher order accurate interpolation is performed. In linear discontinuous and nonlinear discontinuous regions, TVD criterion and MLP limiter are respectively used to remove oscillation. Hence, e-MLP has a number of advantages, as it incorporates useful features of MLP limiter such as multi-dimensional monotonicity and straightforward extensionality to higher order interpolation. It is applicable to local extrema and prevents excessive damping in a linear discontinuous region through application of appropriate limiting criteria. It is efficient because a limiting function is applied only to a discontinuous region. In addition, it is robust against shock instability due to the strict distinction of the computational domain and the use of regional information in a flux scheme as well as a high order interpolation scheme. This new limiting process was applied to numerous test cases including one-dimensional shock/sine wave interaction problem, oblique stationary contact discontinuity, isentropic vortex flow, high speed flow in a blunt body, planar shock/density bubble interaction, shock wave/vortex interaction and, particularly, magnetohydrodynamic (MHD) cloud-shock interaction problems. Through these tests, it was verified that e-MLP substantially enhances the accuracy and efficiency with both continuous and discontinuous multi-dimensional flows.     

An improved algorithm for noise-robust sparse linear prediction of speech

The performance of linear prediction analysis of speech deteriorates rapidly under noisy environments.To tackle this issue,an improved noise-robust sparse linear prediction algorithm is proposed.First,the linear prediction residual of speech is modeled as Student-t distribution,and the additive noise is incorporated explicitly to increase the robustness,thus a probabilistic model for sparse linear prediction of speech is built.Furthermore,variational Bayesian inference is utilized to approximate the intractable posterior distributions of the model parameters,and then the optimal linear prediction parameters are estimated robustly.The experimental results demonstrate the advantage of the developed algorithm in terms of several different metrics compared with the traditional algorithm and the l₁ norm minimization based sparse linear prediction algorithm proposed in recent years.Finally it draws to a conclusion that the proposed algorithm is more robust to noise and is able to increase the speech quality in applications.     

改进的噪声鲁棒语音稀疏线性预测算法

语音线性预测分析算法在噪声环境下性能会急剧恶化,针对这一问题,提出一种改进的噪声鲁棒稀疏线性预测算法。首先采用学生t分布对具有稀疏性的语音线性预测残差建模,并显式考虑加性噪声的影响以提高模型鲁棒性,从而构建完整的概率模型。然后采用变分贝叶斯方法推导模型参数的近似后验分布,最终实现噪声鲁棒的稀疏线性预测参数估计。实验结果表明,与传统算法以及近几年提出的基于l_1范数优化的稀疏线性预测算法相比,该算法在多项指标上具有优势,对环境噪声具有更好的鲁棒性,并且谱失真度更小,因而能够有效提高噪声环境下的语音质量。