摘要
滑动差分倒谱在自动语言辨识的研究中获得了广泛的应用.但是滑动差分倒谱并没有利用语音信号的静态倒谱信息,在方言辨识中的研究表明静态倒谱比差分倒谱含有更多的特征信息.为此,提出了滑动倒谱(SC)的概念,并与滑动差分倒谱特征矢量进行了对比研究.首先利用开发集的语音考察了滑动差分倒谱和滑动倒谱的控制参数在不同取值的情况下对识别性能的影响,利用爬山法确定了这2类特征矢量达到局部最优控制参数组合的路径,然后利用测试集的数据对优化后的2类特征矢量建立的模型进行了闭集辨识和开集辨识.2种情况下的测试结果都表明滑动倒谱的性能优于滑动差分倒谱.并且这2种参数还具有特征互补性,将它们进行决策级数据融合可以进一步提高系统的性能.
Shifted delta cepstra have been widely used in automatic language identification, but only delta cepstrum information is employed. Research on accent identification revealed that detailed cepstrum is more informative than delta cepstrum. So shifted cepstrum was proposed and comparative study was conducted between these two cepstra. Effects of their control parameters on recognition performance were investigated with speech data in the development set. The best paths of these two vectors to reach a locally optimal control parameter combination were determined with the hill-climbing method. Comparative tests performed with speech data both in the closed test set and open test set demonstrated that shifted cepstra is superior to shifted delta cepstra. In addition, they are mutually complementary and data fusion at the decision level could further improve the performance of the system.
出处
《智能系统学报》
2008年第4期336-341,共6页
CAAI Transactions on Intelligent Systems
基金
"十一五"国家863计划重点项目课题(2006AA010102)
关键词
自动语言辨识
滑动倒谱
滑动差分倒谱
高斯混合模型
automatic language identification
shifted cepstra
shifted delta cepstra
Gaussian mixture model
