qiuqiangkong
commented
6 years ago Hi,
Many thanks for your interest! Yes the feature in the paper is log-power, the feature in the code is log spectral. However, they do not have much different because log X^2 = 2 log X. Then we subtract the mean and divide the standard value. So they should be the same.Best wishes,
Qiuqiang
From: gemengsoftware notifications@github.com Sent: 30 August 2018 03:29:50 To: yongxuUSTC/sednn Cc: Subscribed Subject: [yongxuUSTC/sednn] 建议：关于论文中和代码不太一致问题 (#18)
您好， 在读您的论文，An experimental study on speech enhancement based on deep neural networks 和 A regression approach to speech enhancement based on deep neural networks. 发现文中提取特征都是用的 log-power spectral features。而您的python代码实现中是用log spectral features来做的，abs只取了一次。
― You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHubhttps://github.com/yongxuUSTC/sednn/issues/18, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AMt5ye9RRaLYM8aV7wW5Xr9itjsskIvFks5uV04egaJpZM4WSqx-.
您好， 在读您的论文，An experimental study on speech enhancement based on deep neural networks 和 A regression approach to speech enhancement based on deep neural networks. 发现文中提取特征都是用的 log-power spectral features。而您的python代码实现中是用log spectral features来做的，abs只取了一次。