arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1804.04353eess.AScs.AIeess.SPstat.ML
收藏

基于熵和不确定性估计的语音信号全局SNR估计

Global SNR Estimation of Speech Signals using Entropy and Uncertainty Estimates from Dropout Networks

Samsung R&D Institute India, Bangalore(三星印度班加罗尔研发中心)

Rohith Aralikatti, Dilip Margam, Tanay Sharma, Thanda Abhinav, Shankar M Venkatesan

摘要:

本文提出了两种新的方法来估计语音信号的全局信噪比。在这两种方法中,语音识别系统中使用的深度神经网络-隐马尔可夫模型(DNN-HMM)声学模型被用于SNR估计的附加任务。在第一种方法中,计算DNN-HMM输出的熵。最近关于baidu深度学习的研究表明,使用dropout训练的DNN-HMM可以通过将其近似为深度高斯过程来估计模型的不确定性。在第二种方法中,这种近似用于获得模型不确定性估计。噪声特定的回归用于从熵和模型不确定性预测SNR。DNN-HMM在GRID语料库上进行训练,并在来自DEMAND噪声数据库的不同噪声配置文件上进行测试,SNR水平范围为-10 dB至30 dB。

英文摘要:

This paper demonstrates two novel methods to estimate the global SNR of speech signals. In both methods, Deep Neural Network-Hidden Markov Model (DNN-HMM) acoustic model used in speech recognition systems is leveraged for the additional task of SNR estimation. In the first method, the entropy of the DNN-HMM output is computed. Recent work on bayesian deep learning has shown that a DNN-HMM trained with dropout can be used to estimate model uncertainty by approximating it as a deep Gaussian process. In the second method, this approximation is used to obtain model uncertainty estimates. Noise specific regressors are used to predict the SNR from the entropy and model uncertainty. The DNN-HMM is trained on GRID corpus and tested on different noise profiles from the DEMAND noise database at SNR levels ranging from -10 dB to 30 dB.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑