推荐星级:
  • 1
  • 2
  • 3
  • 4
  • 5

为什么深度神经网络适合语音识别

更新时间:2020-03-10 20:02:28 大小:3M 上传用户:xuzhen1查看TA发布的资源 标签:语音识别 下载积分:1分 评价赚积分 (如何评价?) 打赏 收藏 评论(0) 举报

资料介绍

Introduction: ASR History ASR formulation: oGMM/HMM+n-gram +Viterbi search Technical advances(incremental) over past 10 years: oAdaptation(speaker/environment):5% rel. gain oDiscriminative Training:5-10% rel. gain oFeature normalization:5% rel. gain oROVER:5% rel. gain More and more data →better and better accuracy o read speech(>90%), telephony speech(>70%)
omeeting/voicemail recording(<60%)

Neural Network for ASR
1990s: MLP for ASR(Bourlard and Morgan,1994)
o NN/HMM hybrid model(worse than GMM/HMM)
2000s: TANDEM(Hermansky, Elis, et al,2000)
o Use MLP as Feature Extraction (5-10% rel. gain)
2006: DNN for small tasks(Hinton et al,2006)
oRBM-based pre-training for DNN
2010: DNN for small-scale ASR(Mohamed, Yu, et al.2010)
2011: DNN for large-scale ASR oOver 30% rel. gain in Switchboard (Seide et al,2011)

In-house 70-hour Mandarin ASR task; GMM:4000 tied HMM states,30 Gaussians per state DNN:pre-trained;1024 nodes per layer;1-6 hidden layers

部分文件列表

文件名 大小
为什么深度神经网络适合语音识别.pdf 3M

全部评论(0)

暂无评论

上传资源 上传优质资源有赏金

  • 打赏
  • 30日榜单

推荐下载