speech-enhancement

智能算法中的遗传算法实现最优化的文件，对原始的算法有所改进(Intelligent algorithm genetic algorithm to optimize the file to the original algorithm improvements)

电力系统潮流计算程序,利用电力系统潮流计算理论及相关的数学方程中的非线路问题的算法来进行相关的求解。(Power flow calculation in power industry. it easy to use and simulate for power systme.)

基于减法聚类及自适应模糊神经网络的 短期电价预测.(Short-term price forecasting Subtractive Clustering and Adaptive Fuzzy Neural Network.)

采用经典的阴性选择算法N-NSA，用IRIS数据集对算法进行耐受和检测，检测器半径固定不变，检测效率不高，因此需要采用变半径的方法检测(The negative selection algorithm using the classical N-NSA, with the IRIS data set of algorithms for tolerance and detection, the detector radius is fixed, detection efficiency is not high, requiring the use of variable radius method to detect)

MATLAB下用最小二乘设计低通，高通，带通，带阻数字滤波器。(Using MATLAB m-file to design LP, HP, BP and BS fitlers based on Least-square algorithm.)

卷积神经网络是人工神经网络的一种，已成为当前语音分析和图像识别领域的研究热点。它的权值共享网络结构使之更类似于生物神经网络，降低了网络模型的复杂度，减少了权值的数量。自己可已加入数据，和修改梯度。(Convolutional neural network is an artificial neural network, it has become a research hotspot speech analysis and image recognition field. It weights shared network structure to make it more similar to biological neural networks, reduce the complexity of the network model, reducing the number of weights. Own data may have been added, and modify gradients.)

脉冲变压器设计手册..希望对大家有所帮助(Pulse Transformer Design Manual .. hope to be helpful)

《数字信号处理实验指导书（MATLAB 版）》 著者： （美）米特拉著 作译者： 孙洪等译 ISBN号： 7-121-00700-2 出版日期： 2005-01 出版社：电子工业出版社 内有书中个例子的源代程序。MAC，PC及UNIX环境下运行的不同代码。()

说明：  matlab 编写的 守恒-激波-跨声速 喷管 计算源码(matlab program)

圆卷积与线卷积在FIR中的实现，通过程序还开深刻理解两者间的差别和各自优劣。 1.圆卷积用在被截出的某段数据中，此时，长度应该定为和输入数据长度一致（一般输入数据大于hn的数），这时圆卷积结果相当于某段被截断出来的数据以非截断的方式经过滤波而得出的结果；线卷积用在各自独立的，离散的数据中，若截断数据也用线性卷积来计算，那么就相当与这部分数据是与源数据是独立的， 得出的结果时域两端会变长，但频谱基本上一样的。差别可以想象。 2.注意圆卷积最开始的输入数据的次序，构造矩阵时要注意。(Circular convolution with the line convolution in the FIR implementation, the program also opened a deep understanding of the difference between the two and their respective pros and cons. A circular convolution with cut out a certain period of data At this point, the length should be set for the same length as the input data (typically input data greater than hn the number), when the circular convolution is equivalent to a certain the results of the truncated data to the non-truncated filtered line convolution with a separate, discrete data, truncated data with a linear convolution calculation, then the equivalent of this part of the data is independent of the source data, the results of both ends of the time domain becomes long, but the spectrum is basically the same. The difference can imagine. Note that the order of the circular convolution of the input data to construct matrix.)