junhengqi

考虑L的三个不同值：L＝256（3个数据段），L＝128(7个数据段)和L=64（15个数据段）。各自的谱估计图如上图所示。可以明显的看到，加窗明显的减小了频谱上的假谱峰，但也更加进一步平滑了谱峰。所以，对于L＝64的情况，在ω＝0.8π的谱线可以很确定的辨认，但是那两个靠近的谱峰不容易区分。对于L＝128的情况，这种情况提供了在分离和检测间最好的均衡。当然，对于在L＝256时的情况，效果是更好的，能够从谱估计图上明显的分辨出三条谱线的存在以及它们幅度关系的强弱。 除了Welch法外，还可以采用对多个周期图求平均的功率谱估计方法的其他方法如Bartlett法等等，在功率谱估计上也能取得较好的结果。(consider three different values : L = 256 (3 data), L = 128 (7 data) and L = 64 (15 data). Their spectral estimation map as shown in Fig. It is clear to see that increase significantly decreasing window of the spectrum of false peaks, but much more to further smooth the peaks. Therefore, L = 64, in the case = 0.8 spectrum can determine the identification, but that the two peaks around is not easy to distinguish. L = 128 for the situation, which provided the separation and detection among the best balanced. Of course, L = 256 in the case, the effect is better, from the spectral estimation map clearly distinguish between the three lines of the existence and their relationship to the rate of strength. In addition to Welch law, but also can be used to map multiple cycle for the average power)

mel frequency cepstrum coeffcients

demo OceanOptics communication systems.

龙格库塔法的编程,龙格库塔法的编程龙格库塔法的编程(Runge-Kutta method of programming)

实景图像复原处理，各种方法，效果都不错，还有解释说明(Live image restoration processing, a variety of methods, the effect is pretty good)

matlab实例编程一百个源程序，是matlab学习和研究者的很好的参考资料。(matlab example)

Computes the ivnerse value of the conditional BB7 (U given V) copula cdf at a specified point

洗衣机的模糊控制。分3个部分： 1：污泥和油脂隶属函数设计仿真程序； 2：洗涤时间隶属函数设计仿真程序； 3：洗衣机模糊控制系统仿真。(Fuzzy Control washing machine. Of three parts: 1: sludge and grease membership function design simulation program 2: washing time membership function design simulation program 3: Washing fuzzy control system simulation.)

matlabcode for statics information bla bla

The file calculates and plots FM noise sidebands for a carrier. It also does sinusoidal modulation. This simple way of adding noise to a carrier is useful for simulation of PLLs. It turns out, though, that the mean of the randn function is not as close to zero as it could be, and this causes the fft to generate extra sideband energy, which makes it appear as if the noise is not what would be expected. This program gets around this by adding a fudge factor to the randn results to eliminate this problem. It took me ages to figure this out, I hope to reduce a similar effort for others