论文部分内容阅读
由于在说话人识别中梅尔频率滤波器组结构分布不均匀,在低频区域分布密集而在中心频率、高频率分布稀疏,影响了在中、高频段的MEL倒谱系数(MFCC)的提取,本文提出适用于说话人识别的改进MEL滤波器与Mid Mel滤波器相结合得到两种混合特征参数,用此方式来提高中、高频率特征参数提取的精度,从而提高系统识别率。实验结果显示,在同一环境中,新的混合特征参数识别率与识别性能优于传统的特征参数,且运算量较少。
Because of the non-uniform structure of Mel frequency filter bank in speaker recognition, dense distribution in the low frequency region and sparseness in the center frequency and high frequency, it affects the extraction of the MFCC in the middle and high frequency bands, In this paper, we propose two hybrid eigenparameters, which are improved MEL filter and Mid Mel filter, which are suitable for speaker recognition. In this way, the accuracy of feature extraction is improved and the system recognition rate is improved. Experimental results show that in the same environment, the recognition rate and recognition performance of the new hybrid feature parameters are better than the traditional feature parameters, and the computational complexity is low.