Hello,

I'm developing a call-center application using PJSIP (actually, mostly
PJSUA) and currently I'm trying to implement an inband-DTMF detector port
(I can't rely on SIP INFO messages).

Right now, I'm using SpanDSP for reading DTMF digits.  However, it doesn't
seem to work (no digits are received).  Has anyone tried to integrate
SpanDSP and PJSIP before?

Currently, I'm feeding the samples as-is to SpanDSP, and I'm relying on
SpanDSP's callback mechanism to receive the digits:

static pj_status_t dtmfdet_put_frame(pjmedia_port *this_port,
const pjmedia_frame frame)
{
struct dtmfdet_port dport = (struct dtmfdet_port) this_port;
dtmf_rx(&dport->state, (const pj_int16_t) frame->buf,
dport->base.info.samples_per_frame);
return PJ_SUCCESS;
}

static void dtmfdet_digits_rx(void *user_data, const char *digits,
int len)
{
struct dtmfdet_port dport = (struct dtmfdet_port) user_data;
int i;

if (dport->cfg.cb.on_dtmf_digit != NULL) {
  for (i = 0; i < len; i++) {
    dport->cfg.cb.on_dtmf_digit(dport->cfg.call_id, digits[i]);
  }
}

}

Is this the right way to do it?  Should I convert the samples?  I'm
assuming the frame contains linear samples, and as far as I can tell,
SpanDSP expects them to be linear.

Thanks for your help.

Regards,

Maarten

P.S. I've also tried to implement the Goertzel algorithm
(http://en.wikipedia.org/wiki/Goertzel_algorithm) directly, but it didn't
work out, so I thought it would be wise to try integrating SpanDSP first.
Eventually, I'd like to get rid of the SpanDSP dependency, and use Goertzel
directly.

--
Privateer Software Development (www.privateer-software.nl)

• web usability, web accessibility, web development
• cross-platform software development

I don' know about integrating pjsip and spandsp, but why don't you use
rfc2833 or sip info? What codec are you using?

--
Saúl -- "Nunca subestimes el ancho de banda de un camión lleno de disketes."

http://www.saghul.net/

Hi,

Does anyone have an in-band DTMF solution for pjsip?  I know this question
has been raised in the past and I am hoping some crackerjack coder has
figured it out and would like to share it with the group.  I've been seeing
more and more DTMF converted to in-band through various voip providers and I
just not savvy enough to understand how to code this detection in pjsip.

Any help is appreciated.

Sincerely,

Archie

Would someone mind pointing me in the right direction for capturing audio to
a buffer?  I have multiple inbound calls and I would like to check for
in-band DTMF (no rfc2833). What is the API order for setting up PJSIP to get
the audio data?  I just need some framework help -- I have reviewed the
sample files on the web site but I'm having trouble grasping which APIs to
use. I do know how to do play audio files, it's the capturing to a buffer
that has me confused.  I mostly use the PJSUA APIs.

Global

pjsua_pool_create

[Shared pool used for all calls]

Call comes in (for each call do below)

pjmedia_mem_capture_create

[What size buffer should I use? Does it matter?]

[How do I get the call information such clockrate, etc, so I can set this
API to use the same values as the call?  Does it matter? Or should I use the
values in the sample programs?]

pjmedia_mem_capture_set_eof_cb

[How do I connect the call to this media port for so the media port is
"listening" and the callback works]

Call ends

pjmedia_port_destroy

Thank you in advance,

Archie

David,

Thank you for the information!  I appreciate the help.

Take care,

Archie

From: pjsip-bounces@lists.pjsip.org [mailto:pjsip-bounces@lists.pjsip.org]
On Behalf Of David Clark
Sent: Thursday, September 17, 2009 9:43 PM
To: pjsip list; 'pjsip list'
Subject: Re: [pjsip] Capture Call Audio to Buffer

At 10:37 AM 9/15/2009, Archie Rosenblum wrote:

Content-Type: multipart/alternative;
boundary="----=_NextPart_000_0088_01CA35F8.E071DB00"
Content-Language: en-us

Would someone mind pointing me in the right direction for capturing audio to
a buffer?  I have multiple inbound calls and I would like to check for
in-band DTMF (no rfc2833). What is the API order for setting up PJSIP to get
the audio data?  I just need some framework help -- I have reviewed the
sample files on the web site but I'm having trouble grasping which APIs to
use. I do know how to do play audio files, it's the capturing to a buffer
that has me confused.  I mostly use the PJSUA APIs.

Global

pjsua_pool_create
[Shared pool used for all calls]

Call comes in (for each call do below)

pjmedia_mem_capture_create
[What size buffer should I use? Does it matter?]

For FFT analysis I used a buffer size of 2000 bytes.  The idea being the
bigger the buffer size the more accurate the result.  The smaller the buffer
the more real time
the data result.

[How do I get the call information such clockrate, etc, so I can set this
API to use the same values as the call?  Does it matter? Or should I use the
values in the sample programs?]

Yes I think it matters.  I use the pjsua_conf_get_port_info() function to
get this data then pass it along.  It looks something like this:
pjsua_conf_get_port_info(sip_data[line].conf_slot, &info);
sip_data[line].cpa_clock_rate=info.clock_rate;
pjmedia_mem_capture_create(sip_data[line].cpa_pool,
sip_data[line].cpa_int_data, CPA_FFT_SIZE,
info.clock_rate, info.channel_count,
info.samples_per_frame, info.bits_per_sample, 0, &sip_data[line].cpa_port);

pjmedia_mem_capture_set_eof_cb

[How do I connect the call to this media port for so the media port is
"listening" and the callback works]

// line_number is supplied by the application to sort out multiple line
implmentation.
pjmedia_mem_capture_set_eof_cb(sip_data[line].cpa_port,
&sip_data[line].line_number, cpa_fft_got_data);

// add the memory capture port to the bridge
pjsua_conf_add_port(sip_data[line].cpa_pool, sip_data[line].cpa_port,
&sip_data[line].cpa_conf_port);

// then connect the conf_slot for the call and the cpa_conf_slot.
pjsua_conf_connect(sip_data[line].conf_slot, sip_data[line].cpa_conf_port);

The callback function will look like this:
pj_status_t cpa_fft_got_data(pjmedia_port *port, void *usr_data)
{

usr_data will contain the line_number passed int.
Then I think you call pjmedia_port_get_frame(port, frame);
and frame->buf I think has your audio data.

Yea I didn't realize until I typed out this message, my code was missing
that piece.  So
helping you helped me.
}

Call ends

pjmedia_port_destroy

Thank you in advance,
Archie

Visit our blog: http://blog.pjsip.org

pjsip mailing list
pjsip@lists.pjsip.org
http://lists.pjsip.org/mailman/listinfo/pjsip_lists.pjsip.org

This email has been scanned by the MxScan Email Security System.