嵌入时延神经网络的高斯混合模型说话人辨认

提出了一种在高斯混合模型中嵌入时延神经网络的方法。它集成了作为判别性方法的时延神经网络和作为生成性方法的高斯混合模型各自的优点。时延神经网络挖掘了特征向量集的时间信息,并且通过时延网络的变换使需要假设变量独立的最大似然概率（ML）方法更为合理。以最大似然概率为准则,把它们作为一个整体来进行训练。训练过程中,高斯混合模型和神经网络的参数交替更新。实验结果表明,采用所提出的模型在各种信噪比情况下的识别率都比基线系统有所提高,最高能达到21%。

Speaker identification based on GMM with embedded TDNN

This paper proposes a modified Gaussian Mixed Model（GMM） with an embedded Time Delay Neural Network（TDNN）.It integrates the merits of GMM which is generative and TDNN as a Discriminative model.TDNN digests the time information of the feature sets,and through the transformation of the feature vector it makes the hy-pothesis of independence that maximum likelihood needs more reasonable.GMM and TDNN are trained as a whole by means of maximum likelihood.In the process of training,the parameter of GMM and TDNN are updated alternately.Experiments show that the proposed system improves accuracy rate against baseline GMM at all SNR with a maximum to 21%.