search

yongxuUSTC / sednn

deep learning based speech enhancement using keras or pytorch, make it easy to use
http://staff.ustc.edu.cn/~jundu/The%20team/yongxu/demo/SE_DNN_taslp.html
334 stars 124 forks source link

为什么输出y选择7帧语音中间的一帧呢 #33

Open sheepig opened 5 years ago

sheepig commented 5 years ago

训练阶段，x是连续7帧语音，y是中间的一帧，为什么这样设定呢？会对恢复语音精度有影响吗

qiuqiangkong commented 5 years ago

也可以设成1帧，5帧或11帧。帧数更多会捕捉更多上下文信息，但太多了会过拟合，7帧只是一个折中的选择。

Best wishes,

Qiuqiang

From: qyoung notifications@github.com Sent: 27 May 2019 08:15 To: yongxuUSTC/sednn Cc: Subscribed Subject: [yongxuUSTC/sednn] 为什么输出y选择7帧语音中间的一帧呢 (#33)

训练阶段，x是连续7帧语音，y是中间的一帧，为什么这样设定呢？会对恢复语音精度有影响吗

― You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHubhttps://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FyongxuUSTC%2Fsednn%2Fissues%2F33%3Femail_source%3Dnotifications%26email_token%3DADFXTSKQMHT2LZN4VY5NUN3PXODCJA5CNFSM4HPZGYTKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4GV6UX3Q&data=02%7C01%7Cq.kong%40surrey.ac.uk%7C67ec912dd6f0410fa5c808d6e27325c2%7C6b902693107440aa9e21d89446a2ebb5%7C0%7C0%7C636945381515455124&sdata=4ELRl5g5Ze80tX4LmRTRvkOBB%2ByUOeUqVYAlCPovyS0%3D&reserved=0, or mute the threadhttps://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FADFXTSK6XN476HBXOO4ST5DPXODCJANCNFSM4HPZGYTA&data=02%7C01%7Cq.kong%40surrey.ac.uk%7C67ec912dd6f0410fa5c808d6e27325c2%7C6b902693107440aa9e21d89446a2ebb5%7C0%7C0%7C636945381515455124&sdata=5ByhmG%2B2PsVOYxEbotoEaml8s33v7zsFY8KjnrhM%2BxE%3D&reserved=0.