逼近

The new method employs Viterbi algorithm and HMM parameters to produce the best state sequence and then to warp the time within the same state segment using function approximation technique and intersection technique. So a fixed dimensional feature vector can be got. It is then sent to the RBF neural network.

新方法首先利用隐马尔可夫模型参数和Viterbi算法产生语音信号的最佳时间序列，然后利用函数逼近和插值的方法进行时间规正，组成等维的语音识别特征；最后送径向基函数神经网络进行识别。

For getting high fidelity flight simulation, there is a need for a sound system in a flight simulator.

为了使飞行仿真更加逼近真实飞行，给飞行模拟器配置相应的音响系统是十分必要的。

By analyzing the compute results of the integral-exponent-function and the fractional exponent function, conclusion is obtained that the second will improve the precise for its superiority in avoiding "Romberg" phenomenon.

通过对比分析整数次幂和非整数次幂拟合的计算精度，得出了非整数次幂逼近拟合能消除龙格现象，使用非整数次幂进行拟合可使精度提高的结论。

The procedure estimating the regularization parameter is the same as that the adaptive Marquardt algorithm used. The Frechet derivative is directly derived from the forward methods: SAM and FEM. With the theory of SAM and Zhang′s geometric factor, the relationship between the Frechet derivative and the geometric factor is established, at the same time, one useful corollary is given, i.

选取正则化因子时，沿用自适应的Marquardt算法中阻尼因子的选取计算方法；计算Frechet导数时，根据逐次逼近解法和张氏几何因子理论，推导出了Frechet导数与张氏几何因子的关系，得出了Frechet导数为张氏一阶积分几何因子的结论。

Using the method of approximating the Lipschitz functions by the convex functions, we give a sufficient condition for the set where Lipschitz functions on a Hilbert space are Frechet differentiable is a residual set.

我们利用Lipschitz函数空间中凸函数的逼近性质来分析Lipschitz函数的微分性质，得到Hilbert空间上Lipschitz函数的Frechet可微点集是剩余集的一个充分条件，以标题"Lipschitz函数与凸函数之间的一些联系"在中国科学杂志上发表。

In this system,the VVVF(variable voltage variable frequency) control method based on magnetic flux trajectory is applied.In the system,it is the target for the magnetic flux linkage polygon to be greatly approximate to the ideal magnetic flux circle.Considering such limitations as the low speed torque ripple,high speed real time and the working frequency of switching element,through the subsection synchronization control method by applying different numbers of magnetic flux linkage polygon for different frequency range,the perfect precision specifications for speed can be obtained.

系统采用基于磁通轨迹法的他控式变频调速方案，以磁链多边形尽量逼近理想磁通圆为目标，并考虑低速转矩脉动、高速实时性和开关元件的工作频率限制等约束条件，通过采用分段同步控制方式，在不同的频段采用不同的磁链多边形边数，使系统获得了优异的转速精度指标。

For nonlinear friction force in hydraulic position tracking system, partition compensation is to solve the problem of large approximation error caused by non-smooth characteristic of friction when the whole friction is compensated by one fuzzy neural network.

针对液压位置跟踪系统中的非线性摩擦力，提出对摩擦力模型的不同分量分别进行补偿的分部补偿算法，以解决用模糊神经网络对摩擦力整体进行补偿时，因摩擦力非光滑特性引起较大逼近误差的问题。

In order to develop this local property,this paper presents a method,which uses Fourier Series to approximate the boundary shapes of the shafts with periodic boundary of cross section,and uses the conformal mapping in Theory of Function of a Complex Variable to make the complicated boundary shapes of the shafts map a unit circle.Thus elastic theory can analyze...

本文提出一种方法，其实质是：采用富里埃级数来逼近横截面呈周期性变化边界的轴类零件的边界形状；并应用复变函数论中保角映射原理，将零件复杂的边界形状映像成单位圆，从而使弹性理论可以分析和计算复杂边界轴类零件的应力场；再通过留数的求解实现保角映射的反变换。

In theory, any continuous function of n variables can be arbitrarily wellapproximated by the BP Neural Networks.

BP网络从理论上能够逼近任意多变量连续函数，BP网络结构简单，使用方便，是应用最为广泛的一种人工神经网络。

In addition, this network model is a new one that can be widely applied to the fields such as pattern recognition, nonlinear function approximation, time series prediction, and etc...

同时，它也是一种可以广泛应用于模式识别、非线性函数逼近、时间序列预测等领域的新的神经网络模型。

But we don't care about Battlegrounds.

但我们并不在乎沙场中的显露。

Ah! don't mention it, the butcher's shop is a horror.

啊！不用提了。提到肉，真是糟透了。