ACELP | ADPCM | AMBE | AMR | AMR-NB | CELP | Codebook | DTMF | G.723 | G.726 | G.729 | GSM | Interpolation | LPC | LSF | LSP | MELP | PCM | Perceptual | Pitch | PSOL | QCELP | Quantization | SMV | VAD | Vocoder

I have just bought an voice recorder (Speed-Link PDR-3) but it contains just a windows
converter to transform its own recorded voice files into wav/mp3 formats. I would like to use
it in Linux and I'm ready to make an converter just I should know more about the file formats
and the transformation algorithm it uses.

In the windows dll I found just 2 interesting messages:
- ADPCM Encode OK
- unstable LSP

Depending on the selected quality of the recording, the output file is different:
- low quality        file extension is s48 (s4.8 - 480 samples/sec)
- medium quality     file extension is s16 (s16 - 1600 samples/sec)
- hihh quality       file extension is s24 (s24 - 2400 samples/sec) 

I tried to identify and convert it with sox (http://sox.sourceforge.net/) using different adpcm
enconding but without any success. The windows converter tool was made by Sharp, I'm still
trying to get contacted with she. I wanted to get some info about these file formats on the web
but I found nothing.

Is there anybody who can tell me some detail about them? I doubt that Sharp developed an own
file formats/algorithm to store voice content.

Thanks in advance,
Brazso

(You need to be a member of speechcoding -- send a blank email to speechcoding-subscribe@yahoogroups.com )

Hi Phil,

First of all thanks for your answer! To be frankly I know almost nothing 
about DSP technology. I hoped after identifying the used file formats of 
my voice recorder, I can adopt some algorithm, to make a converter in Linux.
Phil Frisbie, Jr. wrote:
> On Tue, March 20, 2007 10:38 am, b...@easymail.hu wrote:
>   
>> I have just bought an voice recorder (Speed-Link PDR-3) but it contains
>> just a windows converter to transform its own recorded voice files into
>> wav/mp3 formats. I would like to use it in Linux and I'm ready to make an
>> converter just I should know more about the file formats and the
>> transformation algorithm it uses.
>>
>> In the windows dll I found just 2 interesting messages:
>> - ADPCM Encode OK
>> - unstable LSP
>>     
>
> Yes, those are interesting but I think your interpretation is wrong.
>
>   
>> Depending on the selected quality of the recording, the output file is
>> different:
>> - low quality        file extension is s48 (s4.8 - 480 samples/sec)
>> - medium quality     file extension is s16 (s16 - 1600 samples/sec)
>> - hihh quality       file extension is s24 (s24 - 2400 samples/sec)
>>     
>
> Do you really mean bytes/sec? That would make more sense.
>   
The dll which is responsible for the conversions contains only the 
following exportable function names:
DESACM1600 DESACM2400 DESACM3200 DESACM480 DESACM640
ENSACM1600 ENSACM2400 ENSACM3200 ENSACM480 ENSACM640
These must be the decoding and encoding functions. Have you heard about 
them? Samples were my tip, perhaps they are bytes as you wrote. I found 
another perhaps interesting phrase in the dll: SUNPLUS SPEECH. All 
information I found about it is in Chinese, but I think the used 
algorithm belongs to this company...

>   
>> I tried to identify and convert it with sox (http://sox.sourceforge.net/)
>> using different adpcm enconding but without any success. The windows
>> converter tool was made by Sharp, I'm still trying to get contacted with
>> she. I wanted to get some info about these file formats on the web but I
>> found nothing.
>>     
>
> Well, I think you are wrong that those files use ADPCM. First, notice the
> the first message says "ADPCM Encode OK", and not "ADPCM Decode OK".
That
> suggests to me that perhaps the wave files are encoded with ADPCM. And the
> other message mentions "unstable LSP", which is referring to Line Spectral
> Pairs, which are NOT used in ADPCM, but in other codecs like CELP.
>   
The windows utility makes the conversion in both directions. At first I 
supposed the encoding message was displayed during wav/mp3 -> 
s48/s16/s24 transformation, but your suggestion makes sense.
>   
>> Is there anybody who can tell me some detail about them? I doubt that
>> Sharp developed an own file formats/algorithm to store voice content.
>>     
>
> When dealing with other than simple codecs (and a codec that uses LSP is
> not simple) you must have more information. There are two many ways they
> could be quantizing those LSP, and even more ways to store the bits into a
> file.
>
>   
I will analyze the result wav file, whether is it ADPCM decoded or not. 
In windows I made a short sample voice record (s16), and I converted it 
into wav, then back to s48/s16/s24 formats so I have enough samples files.

Regards
Brazso

(You need to be a member of speechcoding -- send a blank email to speechcoding-subscribe@yahoogroups.com )

On Tue, March 20, 2007 10:38 am, b...@easymail.hu wrote:
> I have just bought an voice recorder (Speed-Link PDR-3) but it contains
> just a windows converter to transform its own recorded voice files into
> wav/mp3 formats. I would like to use it in Linux and I'm ready to make an
> converter just I should know more about the file formats and the
> transformation algorithm it uses.
>
> In the windows dll I found just 2 interesting messages:
> - ADPCM Encode OK
> - unstable LSP

Yes, those are interesting but I think your interpretation is wrong.

> Depending on the selected quality of the recording, the output file is
> different:
> - low quality        file extension is s48 (s4.8 - 480 samples/sec)
> - medium quality     file extension is s16 (s16 - 1600 samples/sec)
> - hihh quality       file extension is s24 (s24 - 2400 samples/sec)

Do you really mean bytes/sec? That would make more sense.

> I tried to identify and convert it with sox (http://sox.sourceforge.net/)
> using different adpcm enconding but without any success. The windows
> converter tool was made by Sharp, I'm still trying to get contacted with
> she. I wanted to get some info about these file formats on the web but I
> found nothing.

Well, I think you are wrong that those files use ADPCM. First, notice the
the first message says "ADPCM Encode OK", and not "ADPCM Decode OK". That
suggests to me that perhaps the wave files are encoded with ADPCM. And the
other message mentions "unstable LSP", which is referring to Line Spectral
Pairs, which are NOT used in ADPCM, but in other codecs like CELP.

> Is there anybody who can tell me some detail about them? I doubt that
> Sharp developed an own file formats/algorithm to store voice content.

When dealing with other than simple codecs (and a codec that uses LSP is
not simple) you must have more information. There are two many ways they
could be quantizing those LSP, and even more ways to store the bits into a
file.

-- 
Phil Frisbie, Jr.
Hawk Software
http://www.hawksoft.com

(You need to be a member of speechcoding -- send a blank email to speechcoding-subscribe@yahoogroups.com )