你好，请问噪声感知训练部分怎么没有了
qq274943639 opened this issue · 2 comments
qq274943639 commented
你好，请问噪声感知训练部分怎么没有了
qiuqiangkong commented
Hi, which part are you referring?
…________________________________
From: qq274943639 <notifications@github.com>
Sent: 11 February 2020 08:44
To: yongxuUSTC/sednn <sednn@noreply.github.com>
Cc: Subscribed <subscribed@noreply.github.com>
Subject: [yongxuUSTC/sednn] 你好，请问噪声感知训练部分怎么没有了 (#51)
―
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FyongxuUSTC%2Fsednn%2Fissues%2F51%3Femail_source%3Dnotifications%26email_token%3DADFXTSJWPRZPYTKMCK577ITRCJQOHA5CNFSM4KS4PM4KYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4IMPPOBA&data=02%7C01%7Cq.kong%40surrey.ac.uk%7C90d850da560f4eb044b708d7aece96bb%7C6b902693107440aa9e21d89446a2ebb5%7C0%7C0%7C637170074625956730&sdata=S0vwdhM7IatxeHSeNXcV%2BWd4WPM%2F12Yg2xz23aj16%2Bk%3D&reserved=0>, or unsubscribe<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FADFXTSJXEA3FERDMUKADN7DRCJQOHANCNFSM4KS4PM4A&data=02%7C01%7Cq.kong%40surrey.ac.uk%7C90d850da560f4eb044b708d7aece96bb%7C6b902693107440aa9e21d89446a2ebb5%7C0%7C0%7C637170074625961721&sdata=1193JeRF%2BY9lgXTVhiRUZqzhyYOD25XrDdg9xBH1xvs%3D&reserved=0>.
qq274943639 commented
就是在提取特征的时候没有看到有噪声感知训练部分的加进去，不过我看了前面的回复知道了是在带噪语音每一个切片的7帧后面加1帧噪声了谢谢，不过我还有个小问题就是你们用的特征是语音的sft变换实现语音到时频单元的转变，最后一个维度即幅度分量是257，你们为什么不考虑把信号出入伽马通(Gammatone)听觉滤波器来做特征呢？这样只有64个频率分量，我服务器的内存比较小跑的时候经常内存不足，换成这个会不会好点？