《计算机应用研究》|Application Research of Computers

改进批处理RPEM算法用于说话人识别

Improved batch RPEM algorithm for speaker recognition

免费全文下载 (已被下载 次)  
获取PDF全文
作者 项要杰,杨俊安,李晋徽,杨瑞国
机构 1.电子工程学院,合肥 230037;2.安徽省电子制约技术重点实验室,合肥 230037
统计 摘要被查看 次,已被下载
文章编号 1001-3695(2013)12-3579-04
DOI 10.3969/j.issn.1001-3695.2013.12.015
摘要 针对传统EM算法训练GMM不能充分利用训练数据所属高斯分量信息, 从而在一定程度上影响说话人识别性能的缺陷, 采用RPEM (竞争惩罚EM)算法训练GMM, 并引入批处理RPEM算法解决RPEM算法运算量大、收敛速度慢的问题, 同时针对RPEM和批处理RPEM算法训练时方差优化存在的问题进行了改进, 提出了改进的批处理RPEM算法。在Chains 说话人识别数据库上的实验表明, 改进的批处理RPEM算法取得了相对于传统EM、RPEM以及批处理RPEM算法更好的性能, 还极大地提高了训练效率, 减小了运算量, 说明了提出的改进批处理RPEM算法用于说话人识别时的有效性。
关键词 说话人识别;期望最大化算法;竞争惩罚EM算法;批处理竞争惩罚EM算法
基金项目 国家自然科学基金资助项目(60872113)
本文URL http://www.arocmag.com/article/01-2013-12-015.html
英文标题 Improved batch RPEM algorithm for speaker recognition
作者英文名 XIANG Yao-jie, YANG Jun-an, LI Jin-hui, YANG Rui-guo
机构英文名 1. Electronic Engineering Institute, Hefei 230037, China; 2. Key Laboratory of Electronic Restriction, Hefei 230037, China
英文摘要 When the traditional EM algorithm was used to train GMM, it often failed to make full use of the information that the training data belongs to which Gaussian component. And it would influence the performance of speaker recognition to some extent. To solve this problem, this paper adopted the RPEM algorithm to train GMM. But the RPEM algorithm needed a large amount of computation, and its convergence speed was slow. So it introduced the batch RPEM algorithm to overcome the RPEM algorithm's above two shortcomings. However, there were also some problems with the optimization of the variance when it used RPEM or batch RPEM algorithm to train GMM. And it put forward the improved batch RPEM algorithm to solve these problems. The experiments that based on the chains speaker recognition database show that the improved batch RPEM algorithm not only achieves better recognition performance than other three algorithm's recognition performance, but also improves the training efficiency and reduces the amount of computation of the RPEM and batch RPEM.
英文关键词 speaker recognition; expectation maximization(EM)algorithm; rival penalized EM(RPEM) algorithm; batch rival penalized EM(RPEM) algorithm
参考文献 查看稿件参考文献
  [1] KOMLEN D, LOMBAROVIC T, OGRIZEK-TOMAS M, et al. Text independent speaker recognition using LBG vector quantization[C] //Proc of the 34th International Convention. Croatia:MIPRO, 2011:1652-1657.
[2] REYNODLS D, ROSE R. Robust text-independent speaker identification using Gaussian mixture speaker models[J] . IEEE Trans on Speech and Audio Processing, 1995, 3(1):72-83.
[3] CAMPBELL W, CAMPBELL M, GLEASON T J P, et al. Speaker verification using support vector machines and high-level features[J] . IEEE Trans on Audio, Speech, and Language Processing, 2007, 15(7):2085-2094
[4] TAN Jian-ding, TING Hua-nong. Malay speaker identification using neural networks[C] //Proc of International Conference on Information Science and Technology. 2011:476-479.
[5] MINAEI-BIDGOLI B, TOPCHY A, PUNCH W F. A comparison of resampling methods for clustering ensembles[C] //Proc of International Conference on Artificial Intelligence. 2004:939-945.
[6] CHEUNG Y M. Maximum weighted likelihood via rival penalized EM for density mixture clustering with automatic model selection[J] . IEEE Trans on Knowledge and Data Engineering, 2005, 17(6):750-761.
[7] CHEUNG Y M, ZENG Hong. Feature weighted rival penalized EM for Gaussian mixture clustering:automatic feature and model selections in a single paradigm[C] //Proc of International Conference on computational Intelligence and Security. 2007:633-638.
[8] MATZA A, BISTRITZ Y. Speaker recognition with rival penalized EM training[C] //Proc of IEEE International Workshop on Machine Learning for Signal Processing. 2011:1-6.
[9] ZHANG Dan, CHEUNG Y M. A batch rival penalized EM algorithm for Gaussian mixture clustering with automatic model selection[C] //Proc of the 2nd International Conference on Rough Set and Knowledge Technology. Berlin:Springer, 2007:252-259.
[10] CUMMINS F, GRIMALDI M, LEONARD T, et al. The CHAINS corpus:characterizing individual speakers[C] //Proc of SPECOM. 2006:431-435.
收稿日期
修回日期
页码 3579-3582
中图分类号 TP391.4
文献标志码 A