wang.zhiping1@zte.com.cn
2005-Jul-28 17:57 UTC
[Speex-dev] Problems with Speex echo canceller in 1.1.10
Eecho cancellation is used to eliminate the echo which mixed with near end speech, while it was occured by far end speech. It always work like this: Find the similar speech in near end speech and eliminate it. In case 1: Because the there is no echo in near end speech, and the EC module still works on it. So a inverse model is been calculated, and a inverse signal is produced. In this way , you can hear the far end speech. In case 2: Is the NES is produeced by FES? Maybe EC can not recognise the speech as human been just because it is similar. If you want to check the EC, I think you should prepare a echo path model (you can find some typical models on G.168 ). You get a echo signal when the model applied on FES. Once you want a signle talk test, you can use the echo signal as the NES. If you want to test double talk, you can add the echo signal with another speech. ------------------------------ *********************************************** 信息安全声明:本邮件包含信息归ZTE所有, ZTE对该邮件拥有所有权利。请接收者注意 保密,未经发件人书面许可,不得向任何第 三方组织和个人透露本邮件所含信息的全部 或部分。以上声明仅适用于工作邮件。 Information Security Notice: The information contained in this mail is solely property of ZTE Corporation. This mail communication is confidential. Recipients named above are obligated to maintain secrecy and are not permitted to disclose the contents of this communication to others. *********************************************** -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.xiph.org/pipermail/speex-dev/attachments/20050729/471dda2d/attachment.html