/** Create engine channel and source media termination */
mrcp_engine_channel_t* mrcp_engine_source_channel_create(
mrcp_engine_t *engine,
const mrcp_engine_channel_method_vtable_t *channel_vtable,
const mpf_audio_stream_vtable_t *stream_vtable,
void *method_obj,
mpf_codec_descriptor_t *codec_descriptor,
apr_pool_t *pool)
{
mpf_stream_capabilities_t *capabilities;
mpf_audio_stream_t *audio_stream;
mpf_termination_t *termination;
capabilities = mpf_source_stream_capabilities_create(pool);
if(codec_descriptor) {
mpf_codec_capabilities_add(
&capabilities->codecs,
mpf_sample_rate_mask_get(codec_descriptor->sampling_rate),
codec_descriptor->name.buf);
}
else {
mpf_codec_default_capabilities_add(&capabilities->codecs);
}
/* create audio stream */
audio_stream = mpf_audio_stream_create(
method_obj, /* object to associate */
stream_vtable, /* virtual methods table of audio stream */
capabilities, /* stream capabilities */
pool); /* pool to allocate memory from */
if(!audio_stream) {
return NULL;
}
audio_stream->rx_descriptor = codec_descriptor;
/* create media termination */
termination = mpf_raw_termination_create(
NULL, /* no object to associate */
audio_stream, /* audio stream */
NULL, /* no video stream */
pool); /* pool to allocate memory from */
/* create engine channel base */
return mrcp_engine_channel_create(
engine, /* engine */
channel_vtable, /* virtual methods table of engine channel */
method_obj, /* object to associate */
termination, /* media termination, used to terminate audio stream */
pool); /* pool to allocate memory from */
}
RecogChannel* RecogSession::CreateRecogChannel()
{
mrcp_channel_t* pChannel;
mpf_termination_t* pTermination;
mpf_stream_capabilities_t* pCapabilities;
apr_pool_t* pool = GetSessionPool();
/* create channel */
RecogChannel *pRecogChannel = new RecogChannel;
pRecogChannel->m_pMrcpChannel = NULL;
pRecogChannel->m_Streaming = false;
pRecogChannel->m_pAudioIn = NULL;
pRecogChannel->m_TimeToComplete = 0;
/* create source stream capabilities */
pCapabilities = mpf_source_stream_capabilities_create(pool);
GetScenario()->InitCapabilities(pCapabilities);
static const mpf_audio_stream_vtable_t audio_stream_vtable =
{
NULL,
NULL,
NULL,
ReadStream,
NULL,
NULL,
NULL
};
pTermination = CreateAudioTermination(
&audio_stream_vtable, /* virtual methods table of audio stream */
pCapabilities, /* capabilities of audio stream */
pRecogChannel); /* object to associate */
pChannel = CreateMrcpChannel(
MRCP_RECOGNIZER_RESOURCE, /* MRCP resource identifier */
pTermination, /* media termination, used to terminate audio stream */
NULL, /* RTP descriptor, used to create RTP termination (NULL by default) */
pRecogChannel); /* object to associate */
if(!pChannel)
{
delete pRecogChannel;
return NULL;
}
pRecogChannel->m_pMrcpChannel = pChannel;
return pRecogChannel;
}
/** Create demo synthesizer channel derived from engine channel base */
static mrcp_engine_channel_t* mrcp_swift_engine_channel_create(mrcp_engine_t *engine, apr_pool_t *pool)
{
mrcp_swift_engine_t *synth_engine = engine->obj;
mpf_stream_capabilities_t *capabilities;
mpf_termination_t *termination;
mrcp_engine_channel_t *channel;
/* create swift synth channel */
mrcp_swift_channel_t *synth_channel = apr_palloc(pool,sizeof(mrcp_swift_channel_t));
synth_channel->port = NULL;
synth_channel->tts_stream = 0;
synth_channel->channel = NULL;
synth_channel->audio_buffer = NULL;
synth_channel->speak_request = NULL;
synth_channel->stop_response = NULL;
synth_channel->paused = FALSE;
capabilities = mpf_source_stream_capabilities_create(pool);
mpf_codec_capabilities_add(
&capabilities->codecs,
synth_engine->sample_rates,
"LPCM");
/* create media termination */
termination = mrcp_engine_audio_termination_create(
synth_channel, /* object to associate */
&audio_stream_vtable, /* virtual methods table of audio stream */
capabilities, /* stream capabilities */
pool); /* pool to allocate memory from */
/* create engine channel base */
channel = mrcp_engine_channel_create(
engine, /* engine */
&channel_vtable, /* virtual methods table of engine channel */
synth_channel, /* object to associate */
termination, /* associated media termination */
pool); /* pool to allocate memory from */
if(channel) {
synth_channel->audio_buffer = mpf_buffer_create(pool);
}
synth_channel->channel = channel;
return channel;
}
static mpf_termination_t *speech_channel_create_mpf_termination(speech_channel_t *schannel)
{
mpf_termination_t *termination = NULL;
mpf_stream_capabilities_t *capabilities = NULL;
int sample_rates;
if (schannel->type == SPEECH_CHANNEL_SYNTHESIZER)
capabilities = mpf_sink_stream_capabilities_create(schannel->unimrcp_session->pool);
else
capabilities = mpf_source_stream_capabilities_create(schannel->unimrcp_session->pool);
if (capabilities == NULL)
ast_log(LOG_ERROR, "(%s) Unable to create capabilities\n", schannel->name);
/* UniMRCP should transcode whatever the MRCP server wants to use into LPCM
* (host-byte ordered L16) for us. Asterisk may not support all of these.
*/
if (schannel->rate == 16000)
sample_rates = MPF_SAMPLE_RATE_8000 | MPF_SAMPLE_RATE_16000;
else if (schannel->rate == 32000)
sample_rates = MPF_SAMPLE_RATE_8000 | MPF_SAMPLE_RATE_16000 | MPF_SAMPLE_RATE_32000;
else if (schannel->rate == 48000)
sample_rates = MPF_SAMPLE_RATE_8000 | MPF_SAMPLE_RATE_16000 | MPF_SAMPLE_RATE_48000;
else
sample_rates = MPF_SAMPLE_RATE_8000;
/* TO DO : Check if all of these are supported on Asterisk for all codecs. */
if (strcasecmp(schannel->codec, "L16") == 0)
mpf_codec_capabilities_add(&capabilities->codecs, sample_rates, "LPCM");
else
mpf_codec_capabilities_add(&capabilities->codecs, sample_rates, schannel->codec);
termination = mrcp_application_audio_termination_create(
schannel->unimrcp_session, /* Session, termination belongs to. */
&schannel->application->audio_stream_vtable, /* Virtual methods table of audio stream. */
capabilities, /* Capabilities of audio stream. */
schannel); /* Object to associate. */
if (termination == NULL)
ast_log(LOG_ERROR, "(%s) Unable to create termination\n", schannel->name);
return termination;
}
/** Create demo recognizer channel */
static mrcp_channel_t* recog_application_channel_create(mrcp_session_t *session)
{
mrcp_channel_t *channel;
mpf_termination_t *termination;
mpf_stream_capabilities_t *capabilities;
/* create channel */
recog_app_channel_t *recog_channel = apr_palloc(session->pool,sizeof(recog_app_channel_t));
recog_channel->streaming = FALSE;
recog_channel->audio_in = NULL;
recog_channel->time_to_complete = 0;
/* create source stream capabilities */
capabilities = mpf_source_stream_capabilities_create(session->pool);
/* add codec capabilities (Linear PCM) */
mpf_codec_capabilities_add(
&capabilities->codecs,
MPF_SAMPLE_RATE_8000 | MPF_SAMPLE_RATE_16000,
"LPCM");
#if 0
/* more capabilities can be added or replaced */
mpf_codec_capabilities_add(
&capabilities->codecs,
MPF_SAMPLE_RATE_8000 | MPF_SAMPLE_RATE_16000,
"PCMU");
#endif
termination = mrcp_application_audio_termination_create(
session, /* session, termination belongs to */
&audio_stream_vtable, /* virtual methods table of audio stream */
capabilities, /* capabilities of audio stream */
recog_channel); /* object to associate */
channel = mrcp_application_channel_create(
session, /* session, channel belongs to */
MRCP_RECOGNIZER_RESOURCE, /* MRCP resource identifier */
termination, /* media termination, used to terminate audio stream */
NULL, /* RTP descriptor, used to create RTP termination (NULL by default) */
recog_channel); /* object to associate */
return channel;
}
/** Create source media termination */
MRCP_DECLARE(mpf_termination_t*) mrcp_application_source_termination_create(
mrcp_session_t *session,
const mpf_audio_stream_vtable_t *stream_vtable,
mpf_codec_descriptor_t *codec_descriptor,
void *obj)
{
mpf_stream_capabilities_t *capabilities;
mpf_audio_stream_t *audio_stream;
capabilities = mpf_source_stream_capabilities_create(session->pool);
if(codec_descriptor) {
mpf_codec_capabilities_add(
&capabilities->codecs,
mpf_sample_rate_mask_get(codec_descriptor->sampling_rate),
codec_descriptor->name.buf);
}
else {
mpf_codec_default_capabilities_add(&capabilities->codecs);
}
/* create audio stream */
audio_stream = mpf_audio_stream_create(
obj, /* object to associate */
stream_vtable, /* virtual methods table of audio stream */
capabilities, /* stream capabilities */
session->pool); /* memory pool to allocate memory from */
if(!audio_stream) {
return NULL;
}
audio_stream->rx_descriptor = codec_descriptor;
/* create raw termination */
return mpf_raw_termination_create(
NULL, /* no object to associate */
audio_stream, /* audio stream */
NULL, /* no video stream */
session->pool); /* memory pool to allocate memory from */
}
/** Create demo synthesizer channel derived from engine channel base */
static mrcp_engine_channel_t* demo_synth_engine_channel_create(mrcp_engine_t *engine, apr_pool_t *pool)
{
mpf_stream_capabilities_t *capabilities;
mpf_termination_t *termination;
/* create demo synth channel */
demo_synth_channel_t *synth_channel = apr_palloc(pool,sizeof(demo_synth_channel_t));
synth_channel->demo_engine = engine->obj;
synth_channel->speak_request = NULL;
synth_channel->stop_response = NULL;
synth_channel->time_to_complete = 0;
synth_channel->paused = FALSE;
synth_channel->audio_file = NULL;
capabilities = mpf_source_stream_capabilities_create(pool);
mpf_codec_capabilities_add(
&capabilities->codecs,
MPF_SAMPLE_RATE_8000 | MPF_SAMPLE_RATE_16000,
"LPCM");
/* create media termination */
termination = mrcp_engine_audio_termination_create(
synth_channel, /* object to associate */
&audio_stream_vtable, /* virtual methods table of audio stream */
capabilities, /* stream capabilities */
pool); /* pool to allocate memory from */
/* create engine channel base */
synth_channel->channel = mrcp_engine_channel_create(
engine, /* engine */
&channel_vtable, /* virtual methods table of engine channel */
synth_channel, /* object to associate */
termination, /* associated media termination */
pool); /* pool to allocate memory from */
return synth_channel->channel;
}
/** \brief Create recognition channel */
static apt_bool_t uni_recog_channel_create(uni_speech_t *uni_speech, ast_format_compat *format)
{
mrcp_channel_t *channel;
mpf_termination_t *termination;
mpf_stream_capabilities_t *capabilities;
apr_pool_t *pool = mrcp_application_session_pool_get(uni_speech->session);
/* Create source stream capabilities */
capabilities = mpf_source_stream_capabilities_create(pool);
/* Add codec capabilities (Linear PCM) */
mpf_codec_capabilities_add(
&capabilities->codecs,
MPF_SAMPLE_RATE_8000,
"LPCM");
/* Create media termination */
termination = mrcp_application_audio_termination_create(
uni_speech->session, /* session, termination belongs to */
&audio_stream_vtable, /* virtual methods table of audio stream */
capabilities, /* stream capabilities */
uni_speech); /* object to associate */
/* Create MRCP channel */
channel = mrcp_application_channel_create(
uni_speech->session, /* session, channel belongs to */
MRCP_RECOGNIZER_RESOURCE, /* MRCP resource identifier */
termination, /* media termination, used to terminate audio stream */
NULL, /* RTP descriptor, used to create RTP termination (NULL by default) */
uni_speech); /* object to associate */
if(!channel) {
return FALSE;
}
uni_speech->channel = channel;
return TRUE;
}
/** Create ASR session */
ASR_CLIENT_DECLARE(asr_session_t*) asr_session_create(asr_engine_t *engine, const char *profile)
{
mpf_termination_t *termination;
mrcp_channel_t *channel;
mrcp_session_t *session;
const mrcp_app_message_t *app_message;
apr_pool_t *pool;
asr_session_t *asr_session;
mpf_stream_capabilities_t *capabilities;
/* create session */
session = mrcp_application_session_create(engine->mrcp_app,profile,NULL);
if(!session) {
return NULL;
}
pool = mrcp_application_session_pool_get(session);
asr_session = apr_palloc(pool,sizeof(asr_session_t));
mrcp_application_session_object_set(session,asr_session);
/* create source stream capabilities */
capabilities = mpf_source_stream_capabilities_create(pool);
/* add codec capabilities (Linear PCM) */
mpf_codec_capabilities_add(
&capabilities->codecs,
MPF_SAMPLE_RATE_8000,
"LPCM");
termination = mrcp_application_audio_termination_create(
session, /* session, termination belongs to */
&audio_stream_vtable, /* virtual methods table of audio stream */
capabilities, /* capabilities of audio stream */
asr_session); /* object to associate */
channel = mrcp_application_channel_create(
session, /* session, channel belongs to */
MRCP_RECOGNIZER_RESOURCE, /* MRCP resource identifier */
termination, /* media termination, used to terminate audio stream */
NULL, /* RTP descriptor, used to create RTP termination (NULL by default) */
asr_session); /* object to associate */
if(!channel) {
mrcp_application_session_destroy(session);
return NULL;
}
asr_session->engine = engine;
asr_session->mrcp_session = session;
asr_session->mrcp_channel = channel;
asr_session->recog_complete = NULL;
asr_session->input_mode = INPUT_MODE_NONE;
asr_session->streaming = FALSE;
asr_session->audio_in = NULL;
asr_session->media_buffer = NULL;
asr_session->mutex = NULL;
asr_session->wait_object = NULL;
asr_session->app_message = NULL;
/* Create cond wait object and mutex */
apr_thread_mutex_create(&asr_session->mutex,APR_THREAD_MUTEX_DEFAULT,pool);
apr_thread_cond_create(&asr_session->wait_object,pool);
/* Create media buffer */
asr_session->media_buffer = mpf_frame_buffer_create(160,20,pool);
/* Send add channel request and wait for the response */
apr_thread_mutex_lock(asr_session->mutex);
app_message = NULL;
if(mrcp_application_channel_add(asr_session->mrcp_session,asr_session->mrcp_channel) == TRUE) {
apr_thread_cond_wait(asr_session->wait_object,asr_session->mutex);
app_message = asr_session->app_message;
asr_session->app_message = NULL;
}
apr_thread_mutex_unlock(asr_session->mutex);
if(sig_response_check(app_message) == FALSE) {
asr_session_destroy_ex(asr_session,TRUE);
return NULL;
}
return asr_session;
}
