基于通用背景-联合估计(UB-JE)的说话人识别方法

Speaker Recognition Based on Universal Background-Joint Estimation(UB-JE)

在说话人识别中,有效的识别方法是核心.近年来,基于总变化因子分析(i-vector)方法成为了说话人识别领域的主流,其中总变化因子空间的估计是整个算法的关键.本文结合常规的因子分析方法提出一种新的总变化因子空间估计算法,即通用背景–联合估计(Universal background-joint estimation algorithm, UB-JE)算法.首先,根据高斯混合–通用背景模型(Gaussian mixture model-universal background model, GMM-UBM)思想提出总变化矩阵通用背景(UB)算法;其次,根据因子分析理论结合相关文献提出了一种总变化矩阵联合估计(JE)算法;最后,将两种算法相结合得到通用背景–联合估计(UB-JE)算法.采用TIMIT和MDSVC语音数据库,结合i-vector方法将所提的算法与传统算法进行对比实验.结果显示,等错误率(Equal error rate, EER)和最小检测代价函数(Minimum detection cost function, MinDCF)分别提升了8.3%与6.9%,所提方法能够提升i-vector方法的性能.

In the speaker recognition, the effective identification method is the core. In recent years, i-vector method has become the mainstream in the field of speaker recognition, and estimation of the total variation factor space is the key of whole algorithm. In this paper, we propose a new algorithm for total variation factor space estimation named UB-JE,which is combined with conventional factor analysis method. Firstly, the universal background algorithm of total variation matrix is proposed according to Gaussian mixture model-universal background model（GMM-UBM）. Secondly, the joint estimation algorithm of total variation matrix is proposed according to the factor analysis theory and related works.Finally, the two algorithms are combined to get the universal background-joint estimation algorithm（UB-JE）. TIMIT and MDSVC corpus are adopted in the experiment to compare the proposed algorithm with the traditional algorithm.Experimental results show that the equal error rate（EER） and the minimum detection cost function（MinDCF） are improved by 8.3 % and 6.9 %, respectively. The proposed method can improve the performance of i-vector method.