Text only | Text with Images

HydrogenAudio

Lossy Audio Compression => Speech Codecs => Topic started by: hengheng1981 on 2006-02-22 05:32:43

Title: question about the algorithm of AMR-WB
Post by: hengheng1981 on 2006-02-22 05:32:43
Hi,everbody:
  I'm studing document of AMR-WB(G.722.2).I have found that there hasn't the post-processing in decoder. I know the technic of post prosessing which is supposed after reconstructing speech can improve the quality of reconstructed speech .
  In AMR-WB, there just has some process to adaptive codebook ， fixed codebook and gains.
  Dose anybody can help me about why AMR-WB needn't post-processing ?
Title: question about the algorithm of AMR-WB
Post by: QuantumKnot on 2006-02-23 10:08:59
It's quite possible that they didn't put post-processing into AMR-WB standard.  I don't have the technical specs with me at the moment so I'll check them tomorrow.
Title: question about the algorithm of AMR-WB
Post by: hengheng1981 on 2006-02-24 03:04:15
amr-wb is mainly used for wireless communication, right? in addition, for wideband.
    I have read some papars, and found that most narrowband communication use post processing, but for wb, they just adjust the gain in CLEP
    Does it have some relation to bandwidth?
Title: question about the algorithm of AMR-WB
Post by: kwwong on 2006-02-24 03:42:16
Quote
I'm studing document of AMR-WB(G.722.2).I have found that there hasn't the post-processing in decoder. I know the technic of post prosessing which is supposed after reconstructing speech can improve the quality of reconstructed speech .


What post processing technic do you mean?
Title: question about the algorithm of AMR-WB
Post by: hengheng1981 on 2006-02-24 06:04:37
a processing  aims at background noise
Title: question about the algorithm of AMR-WB
Post by: kwwong on 2006-02-26 03:46:15
Quote
a processing  aims at background noise
[a href="index.php?act=findpost&pid=366806"][{POST_SNAPBACK}][/a]

Most probably, you will have no problem in adding additional processing at the decoder.
Title: question about the algorithm of AMR-WB
Post by: QuantumKnot on 2006-02-28 23:12:29
Quote
Quote
I'm studing document of AMR-WB(G.722.2).I have found that there hasn't the post-processing in decoder. I know the technic of post prosessing which is supposed after reconstructing speech can improve the quality of reconstructed speech .


What post processing technic do you mean?
[a href="index.php?act=findpost&pid=366787"][{POST_SNAPBACK}][/a]


Most post-processing in CELP coders involve cascading a long-term and short-term postfilter as well as compensating for gain and reducing spectral tilt.  The short-term postfilter emphasises formant peaks and attenuates spectral valleys between formants, while the long-term postfilter emphasises pitch harmonics.  It makes the reconstructed speech sound better.

As to why they haven't applied it to wideband coding, I'm not too sure.  Perhaps, the existing AMR-WB didn't suffer from the quality problems of narrowband CELP.  (and I seem to have forgotten to look up the tech specs, as promised in my earlier post lol)
Title: question about the algorithm of AMR-WB
Post by: QuantumKnot on 2006-03-02 00:10:27
OK, now browsing the tech specs of AMR-WB decoding.  There is some post-processing in the lower bitrate modes (6.60 and 8.85 kbps) such as anti-sparseness processing, noise enhancement, pitch enhancement, etc.
Text only | Text with Images