SIFT

SPEECH ENCODING USING ADPCM

是关于直方图均衡，以及各种图像方面的增强理论程序，都能运行。也有很详细的说明。很适合初学者(On histogram equalization, and a variety of image enhancements theoretical program to run. There is also a detailed description. It is suitable for beginners )

Optimization Techniques

SSL(Semi-Supervised Learning) with PGP: Implementation of PGP Algorithm introduced in, Fang, Yuan, Kevin Chen-Chuan Chang, and Hady Wirawan Lauw. Graph-based semi-supervised learning: Realizing pointwise smoothness probabilistically. International Conference on Machine Learning (ICML), 2014.

it is a matlab program for ga sa tabu sworm opt heuristics we can also use matlab here. or there can

该代码主要用于求最短路问题，开发环境为Matlab(The code is mainly used for the shortest path problem, the development environment for Matlab)

wsn网络多尺度数据采集于测量 原始matlab仿真代码(Sensor networks, multi-scale measurement)

函数逼近教学例子,可以选择不同训练算法。(Function approximation teaching example, you can choose different training algorithms.)

在径向基函数拟合、径向基函数神经网络和小波神经网络中的几种常用的基函数。包括高斯基函数、逆二次基函数、小波基函数等。(Several basis functions in radial basis function fitting, RBF neural network and wavelet neural networks. Including Gauss function, inverse quadratic basis functions, wavelet function and so on.)

说明：  两种不同的假设： H1 ： 0 xn A fn wn ( ) cos(2 ) ( ) = ++ π θ n=1,2,…,N，f0 为规一化频率 H0 ： xn wn () () = n=1,2,…,N 其中 w[n]是均值为 0，方差为 2 σ n 的高斯白噪声，A 已知，样本间相互 独立，信号与噪声相互独立; 相位θ 是随机变量，它服从均匀分布 1 0 2 ( ) 20 p θ π θ π ?? ≤ ≤ = ??? 其它 1）改变输入信噪比（改变 A 或噪声方差均可），给定虚警概率，画出 输入信噪比与检测概率之间的理论曲线。(注意：理论检测曲线与样本 数有关) 2）改变样本数，用 Monte-Carlo 实验方法得出 PF＝0.001 时输入信噪比 与检测概率之间关系曲线（至少三条），并得出结论。 3）改变 M-C 实验次数，样本数不变，用 Monte-Carlo 实验方法得出 PF ＝0.001 时输入信噪比与检测概率之间关系曲线（至少三条），并得出 结论。(Draw the theoretical curve between input SNR and detection probability)