为了解决传统高斯混合模型(GMM)对初值敏感,在实际训练中极易得到局部最优参数的问题,提出了一种采用微粒群算法优化GMM参数的新方法。
The traditional training methods of Gaussian Mixture Model(GMM) are sensitive to the initial model parameters, which often leads to a local optimal parameter in practice.
仿真实验研究表明:这种新型混合优化算法,计算简单,收敛速度快,初值鲁棒性好。
The simulative experiment demonstrates that this new kind of mixed optimal algorithm is simple in computation, rapid in convergent speed and good in robustness of initial value.
利用混合光学双稳混沌序列的伪随机性和初值敏感性,提出一种用于数字水印中的水印图像置乱算法。
Using the sensitivity to the initial conditions and parameters of mixed-optical bistable chaotic sequence, we have a new scrambling algorithm about watermarking image.
应用推荐