论文部分内容阅读
提出了一种结合全协方差矩阵的RPCL算法，并逐步删除那些只包含少量训练数据的类。这种算法，称之为改进的RPCL算法。它被用于话者识别中优化椭圆基函数网络的结构。实验证明，改进的RPCL算法比原来的RPCL算法能够更好地表征类，并得到更低的误识率。
An RPCL algorithm combined with a full covariance matrix is proposed, and the classes that contain only a small amount of training data are gradually deleted. This algorithm, called the improved RPCL algorithm. It is used to optimize the structure of elliptic basis function networks in speaker recognition. Experiments show that the improved RPCL algorithm can better classify the class and get lower false positive rate than the original RPCL algorithm.