首先,给出结合韵律信息的系统框架。然后,针对汉语的特点,解决了韵律相关的语音识别系统中建模单元选择、模型训练等问题,并在多空间概率分布隐马尔可夫模型(multiple-space distribution hidden Markov mod-el,MSD-HMM)框架下构建了韵...首先,给出结合韵律信息的系统框架。然后,针对汉语的特点,解决了韵律相关的语音识别系统中建模单元选择、模型训练等问题,并在多空间概率分布隐马尔可夫模型(multiple-space distribution hidden Markov mod-el,MSD-HMM)框架下构建了韵律相关的语音识别系统。最后,通过语音识别的实验验证了方法的有效性。在"863"测试集上,该方法能够达到76.18%的带调音节识别正确率。展开更多
In this paper,the authors decide the absolute effect and the Rotation coefficient to scale the improvement of the Sample Rotation to the sample aging, at the same time, analyze the factors to influence the effect of t...In this paper,the authors decide the absolute effect and the Rotation coefficient to scale the improvement of the Sample Rotation to the sample aging, at the same time, analyze the factors to influence the effect of the Sample Rotation.展开更多
文摘首先,给出结合韵律信息的系统框架。然后,针对汉语的特点,解决了韵律相关的语音识别系统中建模单元选择、模型训练等问题,并在多空间概率分布隐马尔可夫模型(multiple-space distribution hidden Markov mod-el,MSD-HMM)框架下构建了韵律相关的语音识别系统。最后,通过语音识别的实验验证了方法的有效性。在"863"测试集上,该方法能够达到76.18%的带调音节识别正确率。
文摘In this paper,the authors decide the absolute effect and the Rotation coefficient to scale the improvement of the Sample Rotation to the sample aging, at the same time, analyze the factors to influence the effect of the Sample Rotation.
