<div>i test your agc fixpoint code ,</div> <div>but it does not work well, far end cant hear anything.</div> <div>printf("%f %f",loudness , st->agc_gain);</div> <div>result is :</div> <div>8.047619 2048.000000<BR>6.285714 280.000000<BR>1.000000 280.000000<BR>2.619048 153.000000<BR>44.523810 2.000000<BR>38.619048 0.000000<BR>47.000000 0.000000<BR>22.809524 0.000000<BR>30.571429 0.000000<BR>17.571429 0.000000<BR>18.380952 0.000000<BR><BR><BR>><STRONG><EM>Thorvald Natvig <thorvald@natvig.com></EM></STRONG> дµÀ£º</div> <BLOCKQUOTE class=replbq style="PADDING-LEFT: 5px; MARGIN-LEFT: 5px; BORDER-LEFT: #1010ff 2px solid">>Hi,<BR>><BR>>Sorry for this being late. Anyway, this has an AGC that works in >fixed <BR>>point. The calculation of the gain is changed slightly even in floating <BR>>point, as loudness_accum is no longer pow()'d, which affects the >initial
<BR>>frames slightly. It gives results in fixed point, but no extensive <BR>>verification has been done, and many of the _ctl functions still do not ><BR>>work, as they need to convert back and forth to decibel scale. I think <BR>>perhaps the best solution there is to have the state struct store the <BR>>decibel value as is done for noise suppression levels etc.<BR>><BR>>-------------- next part --------------<BR>>An embedded and charset-unspecified text was scrubbed...<BR>>Name: speex-agc-temp.diff<BR>>Url: <A href="http://lists.xiph.org/pipermail/speex-dev/attachments/20080522/ea47b677/attachment.txt">http://lists.xiph.org/pipermail/speex-<FONT color=#000000>></FONT>dev/attachments/20080522/ea47b677/attachment.txt</A> <BR></BLOCKQUOTE><BR><p> 
<hr size=1><a href="http://cn.mail.yahoo.com/"> ÑÅ»¢ÓÊÏ䣬ÄúµÄÖÕÉúÓÊÏ䣡</a>