jm [Mon, 12 Feb 2007 15:22:26 +0000 (15:22 +0000)]
more doc
git-svn-id: http://svn.xiph.org/trunk/speex@12458
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 12 Feb 2007 15:16:22 +0000 (15:16 +0000)]
Some formatting so that Mike can stop killing kittens.
git-svn-id: http://svn.xiph.org/trunk/speex@12457
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 12 Feb 2007 14:51:32 +0000 (14:51 +0000)]
updated example for API change.
git-svn-id: http://svn.xiph.org/trunk/speex@12456
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 12 Feb 2007 14:50:06 +0000 (14:50 +0000)]
Separated the normal (integer frequency) calls from the fractional frequency
calls.
git-svn-id: http://svn.xiph.org/trunk/speex@12455
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 12 Feb 2007 14:37:55 +0000 (14:37 +0000)]
Added interleave_int() call.
git-svn-id: http://svn.xiph.org/trunk/speex@12454
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 12 Feb 2007 14:12:59 +0000 (14:12 +0000)]
window type changes with quality.
git-svn-id: http://svn.xiph.org/trunk/speex@12453
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 12 Feb 2007 14:02:32 +0000 (14:02 +0000)]
Tables for Kaiser orders 4, 6, 8.
git-svn-id: http://svn.xiph.org/trunk/speex@12452
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 12 Feb 2007 13:41:59 +0000 (13:41 +0000)]
New quality table, need to implement all orders 4,6,8.
git-svn-id: http://svn.xiph.org/trunk/speex@12451
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 12 Feb 2007 12:50:55 +0000 (12:50 +0000)]
Using a Kaiser window of order 10 instead of the Blackman window. Need to make
the order variable now.
git-svn-id: http://svn.xiph.org/trunk/speex@12450
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Tue, 6 Feb 2007 10:56:43 +0000 (10:56 +0000)]
Making quality map const. Thanks Lennart.
git-svn-id: http://svn.xiph.org/trunk/speex@12439
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Tue, 6 Feb 2007 01:19:26 +0000 (01:19 +0000)]
defining M_PI if it's not.
git-svn-id: http://svn.xiph.org/trunk/speex@12434
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 5 Feb 2007 21:51:07 +0000 (21:51 +0000)]
More API doc
git-svn-id: http://svn.xiph.org/trunk/speex@12431
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 5 Feb 2007 10:58:55 +0000 (10:58 +0000)]
some API documentation
git-svn-id: http://svn.xiph.org/trunk/speex@12429
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 5 Feb 2007 09:20:33 +0000 (09:20 +0000)]
Better (I think) tuning of the filter bandwidth.
git-svn-id: http://svn.xiph.org/trunk/speex@12428
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 5 Feb 2007 02:34:29 +0000 (02:34 +0000)]
filter cutoffs are now quality-dependent
git-svn-id: http://svn.xiph.org/trunk/speex@12427
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sun, 4 Feb 2007 11:55:24 +0000 (11:55 +0000)]
float<->int conversion with a few less bugs (actually works for the first time)
git-svn-id: http://svn.xiph.org/trunk/speex@12425
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sun, 4 Feb 2007 10:48:03 +0000 (10:48 +0000)]
fixed input stride bug.
git-svn-id: http://svn.xiph.org/trunk/speex@12424
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sat, 3 Feb 2007 23:11:37 +0000 (23:11 +0000)]
per-channel data done right (I think).
git-svn-id: http://svn.xiph.org/trunk/speex@12423
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sat, 3 Feb 2007 22:06:09 +0000 (22:06 +0000)]
relecting resampler through a function pointer.
git-svn-id: http://svn.xiph.org/trunk/speex@12422
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sat, 3 Feb 2007 13:31:15 +0000 (13:31 +0000)]
split the processing into two "backend" functions: one for the direct sinc case
and one for the interpolated case.
git-svn-id: http://svn.xiph.org/trunk/speex@12420
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sat, 3 Feb 2007 12:50:57 +0000 (12:50 +0000)]
New, improved. Now with 100% less memory corruption.
git-svn-id: http://svn.xiph.org/trunk/speex@12419
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sat, 3 Feb 2007 12:38:59 +0000 (12:38 +0000)]
Initial support for changing the filter length without screwing up the
output in the process.
git-svn-id: http://svn.xiph.org/trunk/speex@12418
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Fri, 2 Feb 2007 14:29:34 +0000 (14:29 +0000)]
remembering how much was allocated
git-svn-id: http://svn.xiph.org/trunk/speex@12417
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Fri, 2 Feb 2007 14:05:06 +0000 (14:05 +0000)]
oops, removed debug code
git-svn-id: http://svn.xiph.org/trunk/speex@12416
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Fri, 2 Feb 2007 14:03:39 +0000 (14:03 +0000)]
initial support for changing the filter length on the fly...
git-svn-id: http://svn.xiph.org/trunk/speex@12415
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Fri, 2 Feb 2007 10:03:37 +0000 (10:03 +0000)]
Added definitions for quality settings.
git-svn-id: http://svn.xiph.org/trunk/speex@12414
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Thu, 1 Feb 2007 22:44:51 +0000 (22:44 +0000)]
more variable quality stuff
git-svn-id: http://svn.xiph.org/trunk/speex@12412
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Thu, 1 Feb 2007 14:02:40 +0000 (14:02 +0000)]
Changed the resampler API again. Introducing a quality setting.
git-svn-id: http://svn.xiph.org/trunk/speex@12411
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Thu, 1 Feb 2007 06:44:20 +0000 (06:44 +0000)]
oversampling is part of the state.
git-svn-id: http://svn.xiph.org/trunk/speex@12410
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Wed, 31 Jan 2007 06:34:21 +0000 (06:34 +0000)]
split the filter update part.
git-svn-id: http://svn.xiph.org/trunk/speex@12402
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 29 Jan 2007 13:18:50 +0000 (13:18 +0000)]
Some comments. Switched to Blackham window for now (instead of Hanning)
git-svn-id: http://svn.xiph.org/trunk/speex@12388
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Tue, 23 Jan 2007 12:00:08 +0000 (12:00 +0000)]
Added _ctl() calls for the AGC gain limits.
git-svn-id: http://svn.xiph.org/trunk/speex@12382
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Tue, 23 Jan 2007 10:18:23 +0000 (10:18 +0000)]
fixed stupid fixed-point bug. Also defined WORD2INT()
git-svn-id: http://svn.xiph.org/trunk/speex@12381
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Tue, 23 Jan 2007 04:02:54 +0000 (04:02 +0000)]
taking cutoff frequency into account for the gain (oops!!)
git-svn-id: http://svn.xiph.org/trunk/speex@12380
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 22 Jan 2007 15:12:36 +0000 (15:12 +0000)]
resampler inner loop in fixed-point
git-svn-id: http://svn.xiph.org/trunk/speex@12376
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 22 Jan 2007 14:56:23 +0000 (14:56 +0000)]
fixed-point resampler: doing the easy bits first
git-svn-id: http://svn.xiph.org/trunk/speex@12375
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 22 Jan 2007 14:33:23 +0000 (14:33 +0000)]
starting fixed-point version of resampler.
git-svn-id: http://svn.xiph.org/trunk/speex@12374
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 22 Jan 2007 13:53:52 +0000 (13:53 +0000)]
removing the div from the outer loop.
git-svn-id: http://svn.xiph.org/trunk/speex@12373
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 22 Jan 2007 04:58:07 +0000 (04:58 +0000)]
separated into header, source and test program.
git-svn-id: http://svn.xiph.org/trunk/speex@12372
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 22 Jan 2007 03:49:56 +0000 (03:49 +0000)]
interleaved access, fixed short input buf.
git-svn-id: http://svn.xiph.org/trunk/speex@12371
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 22 Jan 2007 00:31:16 +0000 (00:31 +0000)]
implemented reset
git-svn-id: http://svn.xiph.org/trunk/speex@12370
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sun, 21 Jan 2007 21:54:26 +0000 (21:54 +0000)]
changing the sampling rate should now work
git-svn-id: http://svn.xiph.org/trunk/speex@12369
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sun, 21 Jan 2007 13:40:08 +0000 (13:40 +0000)]
API
git-svn-id: http://svn.xiph.org/trunk/speex@12368
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sun, 21 Jan 2007 13:29:28 +0000 (13:29 +0000)]
implemented input/output stride. Not yet tested though.
git-svn-id: http://svn.xiph.org/trunk/speex@12367
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sun, 21 Jan 2007 12:51:25 +0000 (12:51 +0000)]
updated API as discussed with Lennart and Mikko
git-svn-id: http://svn.xiph.org/trunk/speex@12366
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sun, 21 Jan 2007 05:06:57 +0000 (05:06 +0000)]
Implemented cubic interpolation of the (windowed) sinc, reducing the size
of the table.
git-svn-id: http://svn.xiph.org/trunk/speex@12365
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sat, 20 Jan 2007 11:50:09 +0000 (11:50 +0000)]
Down-sampling now doing the right thing and using a lower cutoff frequency.
git-svn-id: http://svn.xiph.org/trunk/speex@12362
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sat, 20 Jan 2007 11:06:55 +0000 (11:06 +0000)]
interpolate after accumulation
git-svn-id: http://svn.xiph.org/trunk/speex@12361
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sat, 20 Jan 2007 09:51:14 +0000 (09:51 +0000)]
Now doing linear interpolation instead of sin(x)/x for large denominators
git-svn-id: http://svn.xiph.org/trunk/speex@12360
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Fri, 19 Jan 2007 03:37:40 +0000 (03:37 +0000)]
API stuff
git-svn-id: http://svn.xiph.org/trunk/speex@12349
0101bb08-14d6-0310-b084-
bc0e0c8e3800
arek [Thu, 18 Jan 2007 13:08:02 +0000 (13:08 +0000)]
Updated/added Xcode build files.
git-svn-id: http://svn.xiph.org/trunk/speex@12345
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Thu, 18 Jan 2007 11:57:03 +0000 (11:57 +0000)]
So doing a lookup is 100x faster than computing sin(x)/x * cos(x/N)... how
surprising! Still need to handle the case where the denominator is too large
to pre-compute everything.
git-svn-id: http://svn.xiph.org/trunk/speex@12344
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Thu, 18 Jan 2007 10:38:06 +0000 (10:38 +0000)]
fixed the alignment of the sinc and removed the if() from the main loop.
Upsampling seems to work good.
git-svn-id: http://svn.xiph.org/trunk/speex@12343
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Thu, 18 Jan 2007 00:33:01 +0000 (00:33 +0000)]
Actually, basic resampling seems to be working...
git-svn-id: http://svn.xiph.org/trunk/speex@12342
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Wed, 17 Jan 2007 21:59:49 +0000 (21:59 +0000)]
output looks almost (but not quite) correct.
git-svn-id: http://svn.xiph.org/trunk/speex@12341
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Wed, 17 Jan 2007 13:38:21 +0000 (13:38 +0000)]
output sucks, but for the first time, I can recognise what went in.
git-svn-id: http://svn.xiph.org/trunk/speex@12336
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Wed, 17 Jan 2007 12:39:13 +0000 (12:39 +0000)]
getting close to non-totally garbage output
git-svn-id: http://svn.xiph.org/trunk/speex@12333
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 15 Jan 2007 21:46:29 +0000 (21:46 +0000)]
...
git-svn-id: http://svn.xiph.org/trunk/speex@12327
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 15 Jan 2007 12:42:15 +0000 (12:42 +0000)]
Some initial resampling code. Nothing even remotely working so far.
git-svn-id: http://svn.xiph.org/trunk/speex@12326
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sun, 14 Jan 2007 13:16:07 +0000 (13:16 +0000)]
minus debug printf()
git-svn-id: http://svn.xiph.org/trunk/speex@12325
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sun, 14 Jan 2007 13:12:28 +0000 (13:12 +0000)]
A bit of tuning and robustness to float overflows.
git-svn-id: http://svn.xiph.org/trunk/speex@12324
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sun, 14 Jan 2007 09:54:06 +0000 (09:54 +0000)]
Some AGC improvements, noise/echo suppression slightly less aggressive by
default.
git-svn-id: http://svn.xiph.org/trunk/speex@12323
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Wed, 10 Jan 2007 06:59:01 +0000 (06:59 +0000)]
Implemented "proportional MDF". Should be more robust to synchronisation
issues.
git-svn-id: http://svn.xiph.org/trunk/speex@12311
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 8 Jan 2007 23:47:31 +0000 (23:47 +0000)]
Some comments on the AEC code.
git-svn-id: http://svn.xiph.org/trunk/speex@12310
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 8 Jan 2007 10:14:35 +0000 (10:14 +0000)]
More file/line debug info
git-svn-id: http://svn.xiph.org/trunk/speex@12309
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 8 Jan 2007 08:58:44 +0000 (08:58 +0000)]
Two-path update decision is now based on an approximation of the power estimator
variance.
git-svn-id: http://svn.xiph.org/trunk/speex@12308
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Thu, 4 Jan 2007 14:40:30 +0000 (14:40 +0000)]
Implemented a dual (foreground + background) filter to improve the robustness
of the AEC. This increases memory use a bit, but I think it's worth it.
git-svn-id: http://svn.xiph.org/trunk/speex@12300
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Wed, 3 Jan 2007 13:37:59 +0000 (13:37 +0000)]
Only the RMS of the excitation is saved in the encoder -- 320 bytes off the
wideband encoder state.
git-svn-id: http://svn.xiph.org/trunk/speex@12294
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Tue, 2 Jan 2007 13:49:38 +0000 (13:49 +0000)]
minor double-codebook simplification for wideband.
git-svn-id: http://svn.xiph.org/trunk/speex@12291
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Tue, 2 Jan 2007 13:37:37 +0000 (13:37 +0000)]
simplified the double_codebook code and improved accuracy slightly
git-svn-id: http://svn.xiph.org/trunk/speex@12290
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Tue, 2 Jan 2007 12:24:26 +0000 (12:24 +0000)]
code shuffling -- no real change.
git-svn-id: http://svn.xiph.org/trunk/speex@12289
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Wed, 27 Dec 2006 13:45:18 +0000 (13:45 +0000)]
fixed-point: making better use of memory -- saved 512 bytes of stack.
git-svn-id: http://svn.xiph.org/trunk/speex@12243
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Tue, 26 Dec 2006 11:37:28 +0000 (11:37 +0000)]
merged some fixed-point and floating-point code in open-loop pitch search.
Also added more debug info for DIV32.
git-svn-id: http://svn.xiph.org/trunk/speex@12242
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 25 Dec 2006 02:31:47 +0000 (02:31 +0000)]
fixed-point: minor tweaks
git-svn-id: http://svn.xiph.org/trunk/speex@12237
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sun, 17 Dec 2006 11:12:58 +0000 (11:12 +0000)]
No more calls to log() in the wideband encoder.
git-svn-id: http://svn.xiph.org/trunk/speex@12225
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sun, 17 Dec 2006 09:47:15 +0000 (09:47 +0000)]
fixed-point: wideband folding gain quantization as a codebook (instead of log)
git-svn-id: http://svn.xiph.org/trunk/speex@12224
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sun, 17 Dec 2006 09:24:49 +0000 (09:24 +0000)]
some comments and a minor tweak
git-svn-id: http://svn.xiph.org/trunk/speex@12223
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sat, 16 Dec 2006 12:57:35 +0000 (12:57 +0000)]
fixed-point: wideband double codebook converted
git-svn-id: http://svn.xiph.org/trunk/speex@12215
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sat, 16 Dec 2006 11:57:40 +0000 (11:57 +0000)]
wideband excitation folding (including gain) now in fixed-point.
git-svn-id: http://svn.xiph.org/trunk/speex@12214
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sat, 16 Dec 2006 08:49:50 +0000 (08:49 +0000)]
more wideband fixed-point work (folding_gain is now in fixed-point).
git-svn-id: http://svn.xiph.org/trunk/speex@12213
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sat, 16 Dec 2006 02:40:27 +0000 (02:40 +0000)]
more wideband fixed-point work
git-svn-id: http://svn.xiph.org/trunk/speex@12212
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Fri, 15 Dec 2006 22:40:50 +0000 (22:40 +0000)]
wideband fixed-point details
git-svn-id: http://svn.xiph.org/trunk/speex@12211
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Wed, 13 Dec 2006 12:53:44 +0000 (12:53 +0000)]
Fixed a very rare overflow problem in the wideband mode, while slightly
increasing accuracy of the fixed-point there. Also, fixed
SPEEX_SET_SUBMODE_ENCODING for wideband (seems like few people were actually
paying attention to that one).
git-svn-id: http://svn.xiph.org/trunk/speex@12208
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Fri, 8 Dec 2006 14:57:20 +0000 (14:57 +0000)]
Removed implementation of all the 32-bit filters now that everything is
computed with 16-bit filters.
git-svn-id: http://svn.xiph.org/trunk/speex@12181
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Fri, 8 Dec 2006 14:43:09 +0000 (14:43 +0000)]
Now using only the 16-bit version of the filters.
git-svn-id: http://svn.xiph.org/trunk/speex@12180
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Fri, 8 Dec 2006 13:40:56 +0000 (13:40 +0000)]
Making use of the wideband encoder input buffer to save memory: 640 bytes off
the encoder state and 640 bytes off the stack requirement.
git-svn-id: http://svn.xiph.org/trunk/speex@12179
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Thu, 7 Dec 2006 11:11:07 +0000 (11:11 +0000)]
temporary kludge to get this to compile
git-svn-id: http://svn.xiph.org/trunk/speex@12177
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Thu, 7 Dec 2006 00:28:40 +0000 (00:28 +0000)]
Make filterbank_psy_smooth more C89-friendly.
git-svn-id: http://svn.xiph.org/trunk/speex@12176
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Wed, 6 Dec 2006 14:55:19 +0000 (14:55 +0000)]
Most of the wideband encoder now in 16-bit. Also, saved another 1.5 kB off the
buffering.
git-svn-id: http://svn.xiph.org/trunk/speex@12174
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Wed, 6 Dec 2006 14:08:39 +0000 (14:08 +0000)]
zero-response now in 16-bit and stored on the stack. About 1 kB saved off the
wideband encoder (including removing unused QMF memory).
git-svn-id: http://svn.xiph.org/trunk/speex@12173
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 4 Dec 2006 13:09:18 +0000 (13:09 +0000)]
No need for an explicit buffer to receive the narrowband innovation.
git-svn-id: http://svn.xiph.org/trunk/speex@12171
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Mon, 4 Dec 2006 11:21:23 +0000 (11:21 +0000)]
Removed 5 kB of useless buffering in the wideband encoder. Also, fixed the
ultra-wideband LPC window.
git-svn-id: http://svn.xiph.org/trunk/speex@12170
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sun, 3 Dec 2006 09:57:08 +0000 (09:57 +0000)]
Another 1 kB off the encoder state by converting the qmf_decomp output to
16-bit.
git-svn-id: http://svn.xiph.org/trunk/speex@12169
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sat, 2 Dec 2006 14:04:43 +0000 (14:04 +0000)]
Fixed a very low-bitrate fixed-point wideband bug (happening mainly with VAD
on) caused by incorrect shift in noise_codebook_quant().
git-svn-id: http://svn.xiph.org/trunk/speex@12165
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Sat, 2 Dec 2006 12:23:05 +0000 (12:23 +0000)]
Got rid of the excitation memory in the decoder and changed the relevant vars
to 16-bit. Last but not least, fixed-point in uwb has been unb0rked.
git-svn-id: http://svn.xiph.org/trunk/speex@12164
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Fri, 1 Dec 2006 14:05:36 +0000 (14:05 +0000)]
Brought EPIC_48K pitch table up-to-date.
git-svn-id: http://svn.xiph.org/trunk/speex@12162
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Fri, 1 Dec 2006 13:33:51 +0000 (13:33 +0000)]
reviving the old EPIC_48K mode.
git-svn-id: http://svn.xiph.org/trunk/speex@12161
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Fri, 1 Dec 2006 09:01:31 +0000 (09:01 +0000)]
Stopped storing zeros in qmf_resynth -- saved ~800 bytes of stack memory in
the wideband decoder
git-svn-id: http://svn.xiph.org/trunk/speex@12160
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Thu, 30 Nov 2006 12:23:37 +0000 (12:23 +0000)]
Saved innovation from nb to wb is now 16-bit precision.
git-svn-id: http://svn.xiph.org/trunk/speex@12158
0101bb08-14d6-0310-b084-
bc0e0c8e3800
jm [Thu, 30 Nov 2006 11:43:11 +0000 (11:43 +0000)]
Bark-domain spectral smoother -- could be useful eventually.
git-svn-id: http://svn.xiph.org/trunk/speex@12156
0101bb08-14d6-0310-b084-
bc0e0c8e3800