app_talkdetect.c

Go to the documentation of this file.

/*
 * Asterisk -- An open source telephony toolkit.
 *
 * Copyright (C) 1999 - 2005, Digium, Inc.
 *
 * Mark Spencer <markster@digium.com>
 *
 * See http://www.asterisk.org for more information about
 * the Asterisk project. Please do not directly contact
 * any of the maintainers of this project for assistance;
 * the project provides a web site, mailing lists and IRC
 * channels for your use.
 *
 * This program is free software, distributed under the terms of
 * the GNU General Public License Version 2. See the LICENSE file
 * at the top of the source tree.
 */

/*! \file
 *
 * \brief Playback a file with audio detect
 *
 * \author Mark Spencer <markster@digium.com>
 * 
 * \ingroup applications
 */

/*** MODULEINFO
   <support_level>extended</support_level>
 ***/

#include "asterisk.h"

ASTERISK_FILE_VERSION(__FILE__, "$Revision: 358907 $")

#include "asterisk/lock.h"
#include "asterisk/file.h"
#include "asterisk/channel.h"
#include "asterisk/pbx.h"
#include "asterisk/module.h"
#include "asterisk/translate.h"
#include "asterisk/utils.h"
#include "asterisk/dsp.h"
#include "asterisk/app.h"

/*** DOCUMENTATION
   <application name="BackgroundDetect" language="en_US">
      <synopsis>
         Background a file with talk detect.
      </synopsis>
      <syntax>
         <parameter name="filename" required="true" />
         <parameter name="sil">
            <para>If not specified, defaults to <literal>1000</literal>.</para>
         </parameter>
         <parameter name="min">
            <para>If not specified, defaults to <literal>100</literal>.</para>
         </parameter>
         <parameter name="max">
            <para>If not specified, defaults to <literal>infinity</literal>.</para>
         </parameter>
         <parameter name="analysistime">
            <para>If not specified, defaults to <literal>infinity</literal>.</para>
         </parameter>
      </syntax>
      <description>
         <para>Plays back <replaceable>filename</replaceable>, waiting for interruption from a given digit (the digit
         must start the beginning of a valid extension, or it will be ignored). During
         the playback of the file, audio is monitored in the receive direction, and if
         a period of non-silence which is greater than <replaceable>min</replaceable> ms yet less than
         <replaceable>max</replaceable> ms is followed by silence for at least <replaceable>sil</replaceable> ms,
         which occurs during the first <replaceable>analysistime</replaceable> ms, then the audio playback is
         aborted and processing jumps to the <replaceable>talk</replaceable> extension, if available.</para>
      </description>
   </application>
 ***/

static char *app = "BackgroundDetect";

static int background_detect_exec(struct ast_channel *chan, const char *data)
{
   int res = 0;
   char *tmp;
   struct ast_frame *fr;
   int notsilent = 0;
   struct timeval start = { 0, 0 };
   struct timeval detection_start = { 0, 0 };
   int sil = 1000;
   int min = 100;
   int max = -1;
   int analysistime = -1;
   int continue_analysis = 1;
   int x;
   struct ast_format origrformat;
   struct ast_dsp *dsp = NULL;
   AST_DECLARE_APP_ARGS(args,
      AST_APP_ARG(filename);
      AST_APP_ARG(silence);
      AST_APP_ARG(min);
      AST_APP_ARG(max);
      AST_APP_ARG(analysistime);
   );

   ast_format_clear(&origrformat);
   if (ast_strlen_zero(data)) {
      ast_log(LOG_WARNING, "BackgroundDetect requires an argument (filename)\n");
      return -1;
   }

   tmp = ast_strdupa(data);
   AST_STANDARD_APP_ARGS(args, tmp);

   if (!ast_strlen_zero(args.silence) && (sscanf(args.silence, "%30d", &x) == 1) && (x > 0)) {
      sil = x;
   }
   if (!ast_strlen_zero(args.min) && (sscanf(args.min, "%30d", &x) == 1) && (x > 0)) {
      min = x;
   }
   if (!ast_strlen_zero(args.max) && (sscanf(args.max, "%30d", &x) == 1) && (x > 0)) {
      max = x;
   }
   if (!ast_strlen_zero(args.analysistime) && (sscanf(args.analysistime, "%30d", &x) == 1) && (x > 0)) {
      analysistime = x;
   }

   ast_debug(1, "Preparing detect of '%s', sil=%d, min=%d, max=%d, analysistime=%d\n", args.filename, sil, min, max, analysistime);
   do {
      if (ast_channel_state(chan) != AST_STATE_UP) {
         if ((res = ast_answer(chan))) {
            break;
         }
      }

      ast_format_copy(&origrformat, ast_channel_readformat(chan));
      if ((ast_set_read_format_by_id(chan, AST_FORMAT_SLINEAR))) {
         ast_log(LOG_WARNING, "Unable to set read format to linear!\n");
         res = -1;
         break;
      }

      if (!(dsp = ast_dsp_new())) {
         ast_log(LOG_WARNING, "Unable to allocate DSP!\n");
         res = -1;
         break;
      }
      ast_stopstream(chan);
      if (ast_streamfile(chan, tmp, ast_channel_language(chan))) {
         ast_log(LOG_WARNING, "ast_streamfile failed on %s for %s\n", ast_channel_name(chan), (char *)data);
         break;
      }
      detection_start = ast_tvnow();
      while (ast_channel_stream(chan)) {
         res = ast_sched_wait(ast_channel_sched(chan));
         if ((res < 0) && !ast_channel_timingfunc(chan)) {
            res = 0;
            break;
         }
         if (res < 0) {
            res = 1000;
         }
         res = ast_waitfor(chan, res);
         if (res < 0) {
            ast_log(LOG_WARNING, "Waitfor failed on %s\n", ast_channel_name(chan));
            break;
         } else if (res > 0) {
            fr = ast_read(chan);
            if (continue_analysis && analysistime >= 0) {
               /* If we have a limit for the time to analyze voice
                * frames and the time has not expired */
               if (ast_tvdiff_ms(ast_tvnow(), detection_start) >= analysistime) {
                  continue_analysis = 0;
                  ast_verb(3, "BackgroundDetect: Talk analysis time complete on %s.\n", ast_channel_name(chan));
               }
            }
            
            if (!fr) {
               res = -1;
               break;
            } else if (fr->frametype == AST_FRAME_DTMF) {
               char t[2];
               t[0] = fr->subclass.integer;
               t[1] = '\0';
               if (ast_canmatch_extension(chan, ast_channel_context(chan), t, 1,
                  S_COR(ast_channel_caller(chan)->id.number.valid, ast_channel_caller(chan)->id.number.str, NULL))) {
                  /* They entered a valid  extension, or might be anyhow */
                  res = fr->subclass.integer;
                  ast_frfree(fr);
                  break;
               }
            } else if ((fr->frametype == AST_FRAME_VOICE) && (fr->subclass.format.id == AST_FORMAT_SLINEAR) && continue_analysis) {
               int totalsilence;
               int ms;
               res = ast_dsp_silence(dsp, fr, &totalsilence);
               if (res && (totalsilence > sil)) {
                  /* We've been quiet a little while */
                  if (notsilent) {
                     /* We had heard some talking */
                     ms = ast_tvdiff_ms(ast_tvnow(), start);
                     ms -= sil;
                     if (ms < 0)
                        ms = 0;
                     if ((ms > min) && ((max < 0) || (ms < max))) {
                        char ms_str[12];
                        ast_debug(1, "Found qualified token of %d ms\n", ms);

                        /* Save detected talk time (in milliseconds) */ 
                        snprintf(ms_str, sizeof(ms_str), "%d", ms);  
                        pbx_builtin_setvar_helper(chan, "TALK_DETECTED", ms_str);

                        ast_goto_if_exists(chan, ast_channel_context(chan), "talk", 1);
                        res = 0;
                        ast_frfree(fr);
                        break;
                     } else {
                        ast_debug(1, "Found unqualified token of %d ms\n", ms);
                     }
                     notsilent = 0;
                  }
               } else {
                  if (!notsilent) {
                     /* Heard some audio, mark the begining of the token */
                     start = ast_tvnow();
                     ast_debug(1, "Start of voice token!\n");
                     notsilent = 1;
                  }
               }
            }
            ast_frfree(fr);
         }
         ast_sched_runq(ast_channel_sched(chan));
      }
      ast_stopstream(chan);
   } while (0);

   if (res > -1) {
      if (origrformat.id && ast_set_read_format(chan, &origrformat)) {
         ast_log(LOG_WARNING, "Failed to restore read format for %s to %s\n", 
            ast_channel_name(chan), ast_getformatname(&origrformat));
      }
   }
   if (dsp) {
      ast_dsp_free(dsp);
   }
   return res;
}

static int unload_module(void)
{
   return ast_unregister_application(app);
}

static int load_module(void)
{
   return ast_register_application_xml(app, background_detect_exec);
}

AST_MODULE_INFO_STANDARD(ASTERISK_GPL_KEY, "Playback with Talk Detection");