opencore-amr

Hi,

On Mon, 14 Nov 2011, Lingfeng Xiong wrote:

> I am trying to encode a voice stream by opencore-amr, this stream is
> captured by PortAudio libray. However, I found whatever data I passed to
> Encoder_Interface_Encode(475), it would always output 13 elements in
> "output" array. So I am wondering how often should I encode my voice stream?

If you encode with mode 0, each AMR frame is encoded to 13 bytes, so that 
looks correct.

> This is the code:
> 
> > import pyaudio
> > import wave
> > import sys
> > import ctypes
> >
> > chunk = 512

One AMR-NB frame is 160 samples (320 for AMR-WB), so you should give it 
that many samples each time you call Encoder_Interface_Encode.

> > FORMAT = pyaudio.paInt8

You should pass 16 bit integer samples to the encoder, this most probably 
is the reason why it doesn't sound right at all.

I'm not familiar with this python wrapper library, so you might need to 
either use 160 or 320 as chunk length, depending on whether it's regarded 
as 320 bytes or 160 16-bit samples.

> > CHANNELS = 1
> > RATE = 22800

AMR-NB normally is 8000 Hz only. You can of course record the data at any 
rate you want, but then you'd have to make sure the playback is at the 
same rate.

> > RECORD_SECONDS = 5
> >
> > p = pyaudio.PyAudio()
> > stream = p.open(format = FORMAT,
> >     channels = CHANNELS,
> >     rate = RATE,
> >     input = True,
> >     frames_per_buffer = chunk)
> >
> > libamr = ctypes.cdll.LoadLibrary("libopencore-amrnb.so")
> > amrstate = libamr.Encoder_Interface_init(0)
> >
> > print "* recording"
> > outfile = open("record.amr", "wb")
> > outfile.write("#!AMR\n")
> > for i in range(0, RATE / chunk * RECORD_SECONDS):
> >     data = stream.read(chunk)
> >     print "data length is %d" % len(data)
> >     outbuf = ctypes.create_string_buffer(600)
> >     n = libamr.Encoder_Interface_Encode(amrstate, 0, data, outbuf, 0);
> >     outfile.write(outbuf[0:n])
> > print "* done recording"
> >
> > outfile.close()
> > stream.close()
> > p.terminate()

// Martin

hi,
Thanks for your kindly help :-)
However, I am still confused about the RATE, frames_per_buffer and 
frame, do you mind giving me some information on them?

----------
Regards,
Lingfeng Xiong

“人民身体之自由应予保障。除现行犯之逮捕由法律另定外，非经司法或警察机关 
依法定程序，不得逮捕拘禁。非由法院依法定程序，不得审问处罚。非依法定程序 
之逮捕、拘禁、审问、处罚，得拒绝之。”
	-- 《中华民国宪法》

On 100/11/14 下午 05:56, Martin Storsjö wrote:
> Hi,
>
> On Mon, 14 Nov 2011, Lingfeng Xiong wrote:
>
>> I am trying to encode a voice stream by opencore-amr, this stream is
>> captured by PortAudio libray. However, I found whatever data I passed to
>> Encoder_Interface_Encode(475), it would always output 13 elements in
>> "output" array. So I am wondering how often should I encode my voice
>> stream?
>
> If you encode with mode 0, each AMR frame is encoded to 13 bytes, so
> that looks correct.
>
>> This is the code:
>>
>> > import pyaudio
>> > import wave
>> > import sys
>> > import ctypes
>> >
>> > chunk = 512
>
> One AMR-NB frame is 160 samples (320 for AMR-WB), so you should give it
> that many samples each time you call Encoder_Interface_Encode.
>
>> > FORMAT = pyaudio.paInt8
>
> You should pass 16 bit integer samples to the encoder, this most
> probably is the reason why it doesn't sound right at all.
>
> I'm not familiar with this python wrapper library, so you might need to
> either use 160 or 320 as chunk length, depending on whether it's
> regarded as 320 bytes or 160 16-bit samples.
>
>> > CHANNELS = 1
>> > RATE = 22800
>
> AMR-NB normally is 8000 Hz only. You can of course record the data at
> any rate you want, but then you'd have to make sure the playback is at
> the same rate.
>
>> > RECORD_SECONDS = 5
>> >
>> > p = pyaudio.PyAudio()
>> > stream = p.open(format = FORMAT,
>> >     channels = CHANNELS,
>> >     rate = RATE,
>> >     input = True,
>> >     frames_per_buffer = chunk)
>> >
>> > libamr = ctypes.cdll.LoadLibrary("libopencore-amrnb.so")
>> > amrstate = libamr.Encoder_Interface_init(0)
>> >
>> > print "* recording"
>> > outfile = open("record.amr", "wb")
>> > outfile.write("#!AMR\n")
>> > for i in range(0, RATE / chunk * RECORD_SECONDS):
>> >     data = stream.read(chunk)
>> >     print "data length is %d" % len(data)
>> >     outbuf = ctypes.create_string_buffer(600)
>> >     n = libamr.Encoder_Interface_Encode(amrstate, 0, data, outbuf, 0);
>> >     outfile.write(outbuf[0:n])
>> > print "* done recording"
>> >
>> > outfile.close()
>> > stream.close()
>> > p.terminate()
>
> // Martin

On Thu, 17 Nov 2011, Lingfeng Xiong wrote:

> However, I am still confused about the RATE, frames_per_buffer and frame, do 
> you mind giving me some information on them?

As I said, the sample rate should be 8000, so you should set RATE to that. 
I'm not sure what frames_per_buffer is in the PyAudio api, but I guess it 
should be 160 here (and chunk should be 160 or 320 depending on how the 
api works, I haven't looked). I don't see any 'frame' parameter in the 
code snippet you posted.

// Martin

hi,
Thanks for your help. However, I am not talking about the parameters 
configuration, I have already adjusted then as you said and it works 
great! :-) What I wanna know is the meaning of this parameter... I am 
confused.. :-(

----------
Regards,
Lingfeng Xiong

"Remember this, take this to heart, live by it, die for it if necessary: 
that our patriotism is medieval, outworn, obsolete; that the modern 
patriotism, the true patriotism, the only rational patriotism, is 
loyalty to the Nation ALL the time, loyalty to the Government when it 
deserves it."
“记住这个，把它记在心里，为它而活，如果需要，为它而死：我们的爱国主义是 
老旧的，废弃的，过时的；现代的爱国主义，真正的爱国主义，理智的爱国主义， 
是永远忠于你的国家，同时在政府值得支持的时候去支持政府。”

	-- Mark Twain

On 100/11/18 下午 06:12, Martin Storsjö wrote:
> On Thu, 17 Nov 2011, Lingfeng Xiong wrote:
>
>> However, I am still confused about the RATE, frames_per_buffer and
>> frame, do you mind giving me some information on them?
>
> As I said, the sample rate should be 8000, so you should set RATE to
> that. I'm not sure what frames_per_buffer is in the PyAudio api, but I
> guess it should be 160 here (and chunk should be 160 or 320 depending on
> how the api works, I haven't looked). I don't see any 'frame' parameter
> in the code snippet you posted.
>
> // Martin
>