[Freetel-codec2] Empty noise trimming causes sound degradation?

Discussion:

Imran

2016-07-19 20:21:33 UTC

Hi All,

I've been experimenting with Codec2 for voice reconstruction. I've
found something odd.

I have a ~2 second long wav file, 8 Khz, 16 bit pcm, mono. I can encode
this file into codec2 no problem.
The first 1.3 seconds is silence. The mic is on, but in a quiet room
with no noise. The wav file spectogram shows a flat line, near 0, with
just a little white noise, for these 1.3 seconds, followed by a quiet
envelope of speech sounds.

So, I encode the wav file into codec2, decode it and playback. Sounds
fine.
Then, I start deleting frames at the head of the encoding. So, I delete
frame 0, then frame 1 .. frame 15. That's about 300 ms of sound I've
removed that has no human sounds in it -- just mic whitenoise.

Then, something weird happens.
While there's still a lot of silence in the playback, when playback
reaches the speech, the voice is noticeably degraded.

I can delete 10 frames, no problem. When I delete more than 10,
degradation starts, and increases as I delete more frames.

Anyone have any hints on why?

Thanks,
Imran

--
Imran
***@fastmail.com

Imran

2016-07-19 21:50:10 UTC

Permalink

More experimentation -- I've decoded the sound to a wav file, trimmed
the samples, played the trimmed wav file, then encoded the wav back into
codec2. The (much shorter) wav file degrades when re-encoded, but plays
back correctly as a wav file.
--
Imran
***@fastmail.com

On Tue, Jul 19, 2016, at 01:39 PM,

Send Freetel-codec2 mailing list submissions to
To subscribe or unsubscribe via the World Wide Web, visit
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
or, via email, send a message with subject or body 'help' to
You can reach the person managing the list at
When replying, please edit your Subject line so it is more specific
than "Re: Contents of Freetel-codec2 digest..."
1. USB interface with codec2 (Faisal Saleem)
2. coding question (Alan Beard)
3. Re: coding question (Brady O'Brien)
4. Re: coding question (Jeroen Vreeken)
5. Empty noise trimming causes sound degradation? (Imran)
----------------------------------------------------------------------
Message: 1
Date: Fri, 15 Jul 2016 07:42:11 +0000 (UTC)
Subject: [Freetel-codec2] USB interface with codec2
Content-Type: text/plain; charset="utf-8"
Hi all,
I am an engineering student and working on a digital voice project. I am
impressed by codec2 and sm1000, the freedv adapter. I just got 2 sm1000
modules and built the firmware sm1000.bin successfully.
My questions are:1) how can i apply codec2 on Recorded voice or Audio
file which is stored in USB? My query is how? i access? from USB a Audio
file.
how can i do?
Thanks in advance for your precious help??Regards,
Faisal Saleem
0092-305-7733854
-------------- next part --------------
An HTML attachment was scrubbed...
------------------------------
Message: 2
Date: Sat, 16 Jul 2016 19:45:41 +1000
Subject: [Freetel-codec2] coding question
Content-Type: text/plain; charset="utf-8"
Hi all,
Can the Codec2 code be split into two sections?
1) From and to the radio audio. To be put into a dedicated DSP device,
see the MMDVM project, Arduino DUE.
2) Human side, microphone and speaker, digital to audio and back, to be
done on a full Linux or Windoze system.
A good project for you Faisal Saleem.
Alan VK2ZIW

Hi all,
I am an engineering student and working on a digital voice project. I am impressed by codec2 and sm1000, the freedv adapter. I just got 2 sm1000 modules and built the firmware sm1000.bin successfully.
1) how can i apply codec2 on Recorded voice or Audio file which is stored in USB? My query is how? i access? from USB a Audio file.
how can i do?
Thanks in advance for your precious help
?
?Regards,
Faisal Saleem
Cell: 0092-305-7733854

Alan
Evil flourishes when good men do nothing.
Consider Jesus.
---------------------------------------------------------------------------
Alan Beard ? ? ? ? ? ? ? Unix Support Technician from 1984 to today
70 Wedmore Rd. ? ? ? ? ? Sun Solaris, AIX, HP/UX, Linux, SCO, MIPS
Emu Heights N.S.W. 2750 ?Routers, terminal servers, printers, terminals
etc..
+61 2 47353013 (h) ? ? ? Support Programming, shell scripting, "C",
assembler
0414 353013 (mobile) ? ? After uni, electronics tech
-------------- next part --------------
An HTML attachment was scrubbed...
------------------------------
Message: 3
Date: Sat, 16 Jul 2016 21:21:20 +1000
Subject: Re: [Freetel-codec2] coding question
Content-Type: text/plain; charset="utf-8"
Hi Alan,
It would be possible to split the modem end of freedv
(fdmdv,cohpsk,fsk,fmfsk) from the codec end. Currently, however, this is
not really supported by the freedv API (freedv_api.c/.h). If anyone is
interested in doing this for any freedv modes, freedv_api.c and
freedv_api.h are good places to look.
Thanks,
Brady O'Brien

*Hi all, Can the Codec2 code be split into two sections? 1) From and to
the radio audio. To be put into a dedicated DSP device, see the MMDVM
project, Arduino DUE. 2) Human side, microphone and speaker, digital to
audio and back, to be done on a full Linux or Windoze system. A good
project for you *
*Faisal Saleem. Alan VK2ZIW On Fri, 15 Jul 2016 07:42:11 +0000 (UTC),
Faisal Saleem wrote*

I am an engineering student and working on a digital voice project. I am

impressed by codec2 and sm1000, the freedv adapter. I just got 2 sm1000
modules and built the firmware sm1000.bin successfully.

1) how can i apply codec2 on Recorded voice or Audio file which is

stored in USB? My query is how i access from USB a Audio file.

how can i do?

Thanks in advance for your precious help

Regards,
*Faisal Saleem*
Cell: 0092-305-7733854

Alan
Evil flourishes when good men do nothing.
Consider Jesus.
---------------------------------------------------------------------------
Alan Beard Unix Support Technician from 1984 to today
70 Wedmore Rd. Sun Solaris, AIX, HP/UX, Linux, SCO, MIPS
Emu Heights N.S.W. 2750 Routers, terminal servers, printers, terminals
etc..
+61 2 47353013 (h) Support Programming, shell scripting, "C",
assembler
0414 353013 (mobile) After uni, electronics tech
------------------------------------------------------------------------------
What NetFlow Analyzer can do for you? Monitors network bandwidth and
traffic
patterns at an interface-level. Reveals which users, apps, and protocols
are
consuming the most bandwidth. Provides multi-vendor support for NetFlow,
J-Flow, sFlow and other flows. Make informed decisions using capacity
planning
reports.http://sdm.link/zohodev2dev
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2

-------------- next part --------------
An HTML attachment was scrubbed...
------------------------------
Message: 4
Date: Sun, 17 Jul 2016 17:53:29 +0200
Subject: Re: [Freetel-codec2] coding question
Content-Type: text/plain; charset=windows-1252

Hi Alan,
It would be possible to split the modem end of freedv
(fdmdv,cohpsk,fsk,fmfsk) from the codec end. Currently, however, this
is not really supported by the freedv API (freedv_api.c/.h). If anyone
is interested in doing this for any freedv modes, freedv_api.c and
freedv_api.h are good places to look.

Actually, I think the codecrx() and codectx()/datatx() functions do
exactly what Alan wants.
Regards,
Jeroen
------------------------------
Message: 5
Date: Tue, 19 Jul 2016 13:21:33 -0700
Subject: [Freetel-codec2] Empty noise trimming causes sound
degradation?
Content-Type: text/plain
Hi All,
I've been experimenting with Codec2 for voice reconstruction. I've
found something odd.
I have a ~2 second long wav file, 8 Khz, 16 bit pcm, mono. I can encode
this file into codec2 no problem.
The first 1.3 seconds is silence. The mic is on, but in a quiet room
with no noise. The wav file spectogram shows a flat line, near 0, with
just a little white noise, for these 1.3 seconds, followed by a quiet
envelope of speech sounds.
So, I encode the wav file into codec2, decode it and playback. Sounds
fine.
Then, I start deleting frames at the head of the encoding. So, I delete
frame 0, then frame 1 .. frame 15. That's about 300 ms of sound I've
removed that has no human sounds in it -- just mic whitenoise.
Then, something weird happens.
While there's still a lot of silence in the playback, when playback
reaches the speech, the voice is noticeably degraded.
I can delete 10 frames, no problem. When I delete more than 10,
degradation starts, and increases as I delete more frames.
Anyone have any hints on why?
Thanks,
Imran
--
Imran
------------------------------
------------------------------------------------------------------------------
What NetFlow Analyzer can do for you? Monitors network bandwidth and
traffic
patterns at an interface-level. Reveals which users, apps, and protocols
are
consuming the most bandwidth. Provides multi-vendor support for NetFlow,
J-Flow, sFlow and other flows. Make informed decisions using capacity
planning
reports.http://sdm.link/zohodev2dev
------------------------------
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
End of Freetel-codec2 Digest, Vol 75, Issue 7
*********************************************

David Rowe

2016-07-20 00:51:51 UTC

Permalink

Hello Imran,

What mode are you using? There isn't much frame-frame memory in Codec 2
(e.g. prediction), in modes like 1300 there is none.

Are you sure the bit-alignment remains correct when you remove frames?

- David

Post by Imran
Hi All,
I've been experimenting with Codec2 for voice reconstruction. I've
found something odd.
I have a ~2 second long wav file, 8 Khz, 16 bit pcm, mono. I can encode
this file into codec2 no problem.
The first 1.3 seconds is silence. The mic is on, but in a quiet room
with no noise. The wav file spectogram shows a flat line, near 0, with
just a little white noise, for these 1.3 seconds, followed by a quiet
envelope of speech sounds.
So, I encode the wav file into codec2, decode it and playback. Sounds
fine.
Then, I start deleting frames at the head of the encoding. So, I delete
frame 0, then frame 1 .. frame 15. That's about 300 ms of sound I've
removed that has no human sounds in it -- just mic whitenoise.
Then, something weird happens.
While there's still a lot of silence in the playback, when playback
reaches the speech, the voice is noticeably degraded.
I can delete 10 frames, no problem. When I delete more than 10,
degradation starts, and increases as I delete more frames.
Anyone have any hints on why?
Thanks,
Imran

Brady O'Brien

2016-07-20 01:35:36 UTC

Permalink

Hello Imran,
Do you have the endianness of the audio going into c2enc/c2dec right? As
written, c2enc and c2dec accept and produce 16 bit signed samples in
whatever the host CPU's endian is; little endian on x86 and most arm chips.
For example, to record and encode some audio on my machine, I'd use the
following commands:

arecord -f S16_LE -r 8k | c2enc 1300 - encoded.c2

note the '-f S16_LE' -- that ensures arecord spits out audio in
little-endian 16 bit signed formed, no wav header.
In the decode direction, I'd use:

c2dec 1300 encoded.c2 - | aplay -f S16_LE -r 8k

To convert from .wav into a format c2enc can consume, I'd use:

sox -t raw -e signed -b 16 -r 8000 speech.wav speech.raw

and from c2dec:

sox -b 16 -e signed -c 1 -r 8000 decoded.raw decoded.wav

Thanks,

Brady O'Brien

Post by David Rowe
Hello Imran,
What mode are you using? There isn't much frame-frame memory in Codec 2
(e.g. prediction), in modes like 1300 there is none.
Are you sure the bit-alignment remains correct when you remove frames?
- David

------------------------------------------------------------------------------
What NetFlow Analyzer can do for you? Monitors network bandwidth and
traffic
patterns at an interface-level. Reveals which users, apps, and protocols
are
consuming the most bandwidth. Provides multi-vendor support for NetFlow,
J-Flow, sFlow and other flows. Make informed decisions using capacity
planning
reports.http://sdm.link/zohodev2dev
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2

Imran

2016-07-20 17:26:43 UTC

Permalink

It does feel like a bitshift error. I'll look deeper into that. Thanks
for the pointer!
--
Imran
***@fastmail.com

On Wed, Jul 20, 2016, at 03:47 AM,

I am an engineering student and working on a digital voice project. I am

impressed by codec2 and sm1000, the freedv adapter. I just got 2 sm1000
modules and built the firmware sm1000.bin successfully.

1) how can i apply codec2 on Recorded voice or Audio file which is

stored in USB? My query is how i access from USB a Audio file.

how can i do?

Thanks in advance for your precious help

Regards,
*Faisal Saleem*
Cell: 0092-305-7733854

------------------------------
Message: 2
Date: Wed, 20 Jul 2016 10:21:51 +0930
Subject: Re: [Freetel-codec2] Empty noise trimming causes sound
degradation?
Content-Type: text/plain; charset=windows-1252; format=flowed
Hello Imran,
What mode are you using? There isn't much frame-frame memory in Codec 2
(e.g. prediction), in modes like 1300 there is none.
Are you sure the bit-alignment remains correct when you remove frames?
- David

Hi All,
I've been experimenting with Codec2 for voice reconstruction. I've
found something odd.
I have a ~2 second long wav file, 8 Khz, 16 bit pcm, mono. I can encode
this file into codec2 no problem.
The first 1.3 seconds is silence. The mic is on, but in a quiet room
with no noise. The wav file spectogram shows a flat line, near 0, with
just a little white noise, for these 1.3 seconds, followed by a quiet
envelope of speech sounds.
So, I encode the wav file into codec2, decode it and playback. Sounds
fine.
Then, I start deleting frames at the head of the encoding. So, I delete
frame 0, then frame 1 .. frame 15. That's about 300 ms of sound I've
removed that has no human sounds in it -- just mic whitenoise.
Then, something weird happens.
While there's still a lot of silence in the playback, when playback
reaches the speech, the voice is noticeably degraded.
I can delete 10 frames, no problem. When I delete more than 10,
degradation starts, and increases as I delete more frames.
Anyone have any hints on why?
Thanks,
Imran

------------------------------
Message: 3
Date: Tue, 19 Jul 2016 20:35:36 -0500
Subject: Re: [Freetel-codec2] Empty noise trimming causes sound
degradation?
Content-Type: text/plain; charset="utf-8"
Hello Imran,
Do you have the endianness of the audio going into c2enc/c2dec right? As
written, c2enc and c2dec accept and produce 16 bit signed samples in
whatever the host CPU's endian is; little endian on x86 and most arm
chips.
For example, to record and encode some audio on my machine, I'd use the
arecord -f S16_LE -r 8k | c2enc 1300 - encoded.c2
note the '-f S16_LE' -- that ensures arecord spits out audio in
little-endian 16 bit signed formed, no wav header.
c2dec 1300 encoded.c2 - | aplay -f S16_LE -r 8k
sox -t raw -e signed -b 16 -r 8000 speech.wav speech.raw
sox -b 16 -e signed -c 1 -r 8000 decoded.raw decoded.wav
Thanks,
Brady O'Brien

Hello Imran,
What mode are you using? There isn't much frame-frame memory in Codec 2
(e.g. prediction), in modes like 1300 there is none.
Are you sure the bit-alignment remains correct when you remove frames?
- David

Hi All,
I've been experimenting with Codec2 for voice reconstruction. I've
found something odd.
I have a ~2 second long wav file, 8 Khz, 16 bit pcm, mono. I can encode
this file into codec2 no problem.
The first 1.3 seconds is silence. The mic is on, but in a quiet room
with no noise. The wav file spectogram shows a flat line, near 0, with
just a little white noise, for these 1.3 seconds, followed by a quiet
envelope of speech sounds.
So, I encode the wav file into codec2, decode it and playback. Sounds
fine.
Then, I start deleting frames at the head of the encoding. So, I delete
frame 0, then frame 1 .. frame 15. That's about 300 ms of sound I've
removed that has no human sounds in it -- just mic whitenoise.
Then, something weird happens.
While there's still a lot of silence in the playback, when playback
reaches the speech, the voice is noticeably degraded.
I can delete 10 frames, no problem. When I delete more than 10,
degradation starts, and increases as I delete more frames.
Anyone have any hints on why?
Thanks,
Imran

-------------- next part --------------
An HTML attachment was scrubbed...
------------------------------
Message: 4
Date: Wed, 20 Jul 2016 13:46:15 +0300
Subject: [Freetel-codec2] Codec2 for STM32
Content-Type: text/plain; charset="utf-8"
Hello!
I want to use Codec2 for STM32 (I want to try it on Discovery board).
I? downloaded source texts from website, but I don't understand what
files need to be included to my project. Help please!
I use CubeMX and HAL libraries from ST, a development environment - IAR.
I ask to excuse me for my English.
Thanks!
--
Marat Galyamov
-------------- next part --------------
An HTML attachment was scrubbed...
------------------------------
------------------------------------------------------------------------------
What NetFlow Analyzer can do for you? Monitors network bandwidth and
traffic
patterns at an interface-level. Reveals which users, apps, and protocols
are
consuming the most bandwidth. Provides multi-vendor support for NetFlow,
J-Flow, sFlow and other flows. Make informed decisions using capacity
planning
reports.http://sdm.link/zohodev2dev
------------------------------
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
End of Freetel-codec2 Digest, Vol 75, Issue 8
*********************************************