资 源 简 介
在实时平台上,高斯混合模型(GMM)具有计算有效性和易于实现的优点。最大似然规则中,模型参数不
断更新,但由于爬山特征,任意的原始模型参数估计通常将导致局部最优 遗传算法(GA)适于求解复杂组合优化问
题及非线性函数优化。提出了基于说话人识别的可以解决GMM局部最优问题的GMM/GA新算法,实验结果表明,
提出的GMM/GA新算法比纯粹的GMM算法能获得更优的效果。
- In real-time platform, the Gaussian mixture model (GMM) with the calculation of the effectiveness and easy to realize benefits. Maximum likelihood rule, the model parameters are not
Broken updates, but due to climbing features, any of the original model parameter estimation will usually result in local optimum genetic algorithm (GA) is suitable for solving complex combinatorial optimization question
Title and non-linear function optimization. Proposed speaker recognition based on GMM can solve the problem of local optimal GMM/GA new algorithm, experimental results show that the
Proposed GMM/GA new algorithm than purely GMM algorithm can get better results.
