ben.carrasco/freeswitch
1
0
Fork 0
mirror of https://github.com/signalwire/freeswitch.git synced 2026-07-24 05:02:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

[Core] Add partial Asr support

Browse Source
This commit is contained in:
Seven Du
2020-02-13 00:34:59 +08:00
committed by Andrey Volk
parent b1714199ce
commit 623fecd599
5 changed files with 181 additions and 47 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/mod/applications/mod_test/mod_test.c
+11 -2
View File
@@ -68,6 +68,7 @@ typedef struct {
char *grammar;
char *channel_uuid;
switch_vad_t *vad;
int partial;
} test_asr_t;
@@ -268,12 +269,17 @@ static switch_status_t test_asr_get_results(switch_asr_handle_t *ah, char **resu
}
if (switch_test_flag(context, ASRFLAG_RESULT)) {
int is_partial = context->partial-- > 0 ? 1 : 0;
*resultstr = switch_mprintf("{\"grammar\": \"%s\", \"text\": \"%s\", \"confidence\": %f}", context->grammar, context->result_text, context->result_confidence);
switch_log_printf(SWITCH_CHANNEL_UUID_LOG(context->channel_uuid), SWITCH_LOG_ERROR, "Result: %s\n", *resultstr);
switch_log_printf(SWITCH_CHANNEL_UUID_LOG(context->channel_uuid), SWITCH_LOG_NOTICE, "%sResult: %s\n", is_partial ? "Partial " : "Final ", *resultstr);
status = SWITCH_STATUS_SUCCESS;
if (is_partial) {
status = SWITCH_STATUS_MORE_DATA;
} else {
status = SWITCH_STATUS_SUCCESS;
}
} else if (switch_test_flag(context, ASRFLAG_NOINPUT_TIMEOUT)) {
switch_log_printf(SWITCH_CHANNEL_UUID_LOG(context->channel_uuid), SWITCH_LOG_DEBUG, "Result: NO INPUT\n");
@@ -361,6 +367,9 @@ static void test_asr_text_param(switch_asr_handle_t *ah, char *param, const char
} else if (!strcasecmp("confidence", param) && fval >= 0.0) {
context->result_confidence = fval;
switch_log_printf(SWITCH_CHANNEL_UUID_LOG(context->channel_uuid), SWITCH_LOG_DEBUG, "confidence = %f\n", fval);
} else if (!strcasecmp("partial", param) && switch_true(val)) {
context->partial = 3;
switch_log_printf(SWITCH_CHANNEL_UUID_LOG(context->channel_uuid), SWITCH_LOG_DEBUG, "partial = %d\n", context->partial);
}
}
}
src/switch_cpp.cpp
+3 -3
View File
@@ -1055,7 +1055,7 @@ SWITCH_DECLARE(char *) CoreSession::playAndDetectSpeech(char *file, char *engine
char *result = NULL;
switch_status_t status = switch_ivr_play_and_detect_speech(session, file, engine, grammar, &result, 0, NULL);
switch_status_t status = switch_ivr_play_and_detect_speech(session, file, engine, grammar, &result, 0, ap);
if (status == SWITCH_STATUS_SUCCESS) {
// good
} else if (status == SWITCH_STATUS_GENERR) {
@@ -1063,12 +1063,12 @@ SWITCH_DECLARE(char *) CoreSession::playAndDetectSpeech(char *file, char *engine
} else if (status == SWITCH_STATUS_NOT_INITALIZED) {
switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_ERROR, "ASR INIT ERROR\n");
} else {
switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_ERROR, "ERROR\n");
switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_ERROR, "ERROR status = %d\n", status);
}
end_allow_threads();
return result; // remeber to free me
return result ? strdup(result) : NULL; // remeber to free me
}
SWITCH_DECLARE(void) CoreSession::say(const char *tosay, const char *module_name, const char *say_type, const char *say_method, const char *say_gender)
src/switch_ivr_async.c
+31 -14
View File
@@ -4696,8 +4696,16 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_bind_dtmf_meta_session(switch_core_se
typedef struct {
int done;
char *result;
switch_input_args_t *original_args;
} play_and_detect_speech_state_t;
static void deliver_asr_event(switch_core_session_t *session, switch_event_t *event, switch_input_args_t *args)
{
if (args && args->input_callback) {
args->input_callback(session, (void *)event, SWITCH_INPUT_TYPE_EVENT, args->buf, args->buflen);
}
}
static switch_status_t play_and_detect_input_callback(switch_core_session_t *session, void *input, switch_input_type_t input_type, void *data, unsigned int len)
{
play_and_detect_speech_state_t *state = (play_and_detect_speech_state_t *)data;
@@ -4708,7 +4716,10 @@ static switch_status_t play_and_detect_input_callback(switch_core_session_t *ses
event = (switch_event_t *)input;
if (event->event_id == SWITCH_EVENT_DETECTED_SPEECH) {
const char *speech_type = switch_event_get_header(event, "Speech-Type");
if (!zstr(speech_type)) {
deliver_asr_event(session, event, state->original_args);
if (!strcasecmp(speech_type, "detected-speech")) {
const char *result;
switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_INFO, "(%s) DETECTED SPEECH\n", switch_channel_get_name(channel));
@@ -4718,8 +4729,11 @@ static switch_status_t play_and_detect_input_callback(switch_core_session_t *ses
} else {
state->result = "";
}
state->original_args = NULL;
state->done = PLAY_AND_DETECT_DONE_RECOGNIZING;
return SWITCH_STATUS_BREAK;
} else if (!strcasecmp(speech_type, "detected-partial-speech")) {
// ok
} else if (!strcasecmp(speech_type, "begin-speaking")) {
switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_INFO, "(%s) START OF SPEECH\n", switch_channel_get_name(channel));
return SWITCH_STATUS_BREAK;
@@ -4727,6 +4741,8 @@ static switch_status_t play_and_detect_input_callback(switch_core_session_t *ses
state->done = PLAY_AND_DETECT_DONE_RECOGNIZING;
state->result = "";
return SWITCH_STATUS_BREAK;
} else {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "unhandled speech type %s\n", speech_type);
}
}
}
@@ -4765,7 +4781,7 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_play_and_detect_speech(switch_core_se
switch_status_t status = SWITCH_STATUS_FALSE;
int recognizing = 0;
switch_input_args_t myargs = { 0 };
play_and_detect_speech_state_t state = { 0, "" };
play_and_detect_speech_state_t state = { 0, "", NULL };
switch_channel_t *channel = switch_core_session_get_channel(session);
arg_recursion_check_start(args);
@@ -4776,10 +4792,6 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_play_and_detect_speech(switch_core_se
if (!input_timeout) input_timeout = 5000;
if (!args) {
args = &myargs;
}
/* start speech detection */
if ((status = switch_ivr_detect_speech(session, mod_name, grammar, "", NULL, NULL)) != SWITCH_STATUS_SUCCESS) {
/* map SWITCH_STATUS_FALSE to SWITCH_STATUS_GENERR to indicate grammar load failed
@@ -4792,12 +4804,19 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_play_and_detect_speech(switch_core_se
recognizing = 1;
/* play the prompt, looking for detection result */
args->input_callback = play_and_detect_input_callback;
args->buf = &state;
args->buflen = sizeof(state);
status = switch_ivr_play_file(session, NULL, file, args);
if (args->dmachine && switch_ivr_dmachine_last_ping(args->dmachine) != SWITCH_STATUS_SUCCESS) {
if (args) {
state.original_args = args;
myargs.dmachine = args->dmachine;
}
myargs.input_callback = play_and_detect_input_callback;
myargs.buf = &state;
myargs.buflen = sizeof(state);
status = switch_ivr_play_file(session, NULL, file, &myargs);
if (args && args->dmachine && switch_ivr_dmachine_last_ping(args->dmachine) != SWITCH_STATUS_SUCCESS) {
state.done |= PLAY_AND_DETECT_DONE;
goto done;
}
@@ -4812,9 +4831,9 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_play_and_detect_speech(switch_core_se
switch_ivr_detect_speech_start_input_timers(session);
switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_INFO, "(%s) WAITING FOR RESULT\n", switch_channel_get_name(channel));
while (!state.done && switch_channel_ready(channel)) {
status = switch_ivr_sleep(session, input_timeout, SWITCH_FALSE, args);
status = switch_ivr_sleep(session, input_timeout, SWITCH_FALSE, &myargs);
if (args->dmachine && switch_ivr_dmachine_last_ping(args->dmachine) != SWITCH_STATUS_SUCCESS) {
if (args && args->dmachine && switch_ivr_dmachine_last_ping(args->dmachine) != SWITCH_STATUS_SUCCESS) {
state.done |= PLAY_AND_DETECT_DONE;
goto done;
}
@@ -4826,8 +4845,6 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_play_and_detect_speech(switch_core_se
}
}
done:
if (recognizing && !(state.done & PLAY_AND_DETECT_DONE_RECOGNIZING)) {
switch_ivr_pause_detect_speech(session);
src/switch_ivr_play_say.c
+31 -18
View File
@@ -3192,8 +3192,16 @@ typedef struct {
char terminator;
switch_time_t last_digit_time;
switch_bool_t is_speech;
switch_input_args_t *original_args;
} switch_collect_input_state_t;
static void deliver_asr_event(switch_core_session_t *session, switch_event_t *event, switch_input_args_t *args)
{
if (args && args->input_callback) {
args->input_callback(session, (void *)event, SWITCH_INPUT_TYPE_EVENT, args->buf, args->buflen);
}
}
static switch_status_t switch_collect_input_callback(switch_core_session_t *session, void *input, switch_input_type_t input_type, void *data, unsigned int len)
{
switch_collect_input_state_t *state = (switch_collect_input_state_t *)data;
@@ -3209,13 +3217,15 @@ static switch_status_t switch_collect_input_callback(switch_core_session_t *sess
if (zstr(speech_type)) return SWITCH_STATUS_SUCCESS;
deliver_asr_event(session, event, state->original_args);
if (!strcasecmp(speech_type, "detected-speech")) {
const char *result = switch_event_get_body(event);
/* stop waiting for speech */
switch_set_flag(state, SWITCH_COLLECT_INPUT_SPEECH_DONE);
switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_INFO, "(%s) DETECTED SPEECH\n", switch_channel_get_name(channel));
switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_INFO, "(%s) DETECTED SPEECH %s\n", switch_channel_get_name(channel), speech_type);
if (!zstr(result)) {
state->recognition_result = cJSON_Parse(result);
@@ -3230,15 +3240,12 @@ static switch_status_t switch_collect_input_callback(switch_core_session_t *sess
}
}
return SWITCH_STATUS_BREAK;
}
if (!strcasecmp("closed", speech_type)) {
} else if (!strcasecmp(speech_type, "detected-partial-speech")) {
} else if (!strcasecmp("closed", speech_type)) {
/* stop waiting for speech */
switch_set_flag(state, SWITCH_COLLECT_INPUT_SPEECH_DONE);
return SWITCH_STATUS_BREAK;
}
if (!strcasecmp(speech_type, "begin-speaking")) {
} else if (!strcasecmp(speech_type, "begin-speaking")) {
switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_INFO, "(%s) START OF SPEECH\n", switch_channel_get_name(channel));
state->is_speech = SWITCH_TRUE;
@@ -3246,6 +3253,8 @@ static switch_status_t switch_collect_input_callback(switch_core_session_t *sess
/* barge in on prompt */
return SWITCH_STATUS_BREAK;
}
} else {
switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_INFO, "Unhandled Speech-Type %s\n", speech_type);
}
}
@@ -3360,10 +3369,6 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_play_and_collect_input(switch_core_se
}
}
if (!args) {
args = &myargs;
}
/* start speech recognition, if enabled */
if (recognizer_grammar && recognizer_mod_name) {
if ((status = switch_ivr_detect_speech(session, recognizer_mod_name, recognizer_grammar, "", NULL, NULL)) != SWITCH_STATUS_SUCCESS) {
@@ -3380,14 +3385,22 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_play_and_collect_input(switch_core_se
}
/* play the prompt, looking for input result */
args->input_callback = switch_collect_input_callback;
args->buf = &state;
args->buflen = sizeof(state);
if (args) {
state.original_args = args;
myargs.dmachine = args->dmachine;
}
myargs.input_callback = switch_collect_input_callback;
myargs.buf = &state;
myargs.buflen = sizeof(state);
switch_set_flag(&state, SWITCH_COLLECT_INPUT_PROMPT);
status = switch_ivr_play_file(session, NULL, prompt, args);
status = switch_ivr_play_file(session, NULL, prompt, &myargs);
switch_clear_flag(&state, SWITCH_COLLECT_INPUT_PROMPT);
if (args->dmachine && switch_ivr_dmachine_last_ping(args->dmachine) != SWITCH_STATUS_SUCCESS) {
if (args && args->dmachine && switch_ivr_dmachine_last_ping(args->dmachine) != SWITCH_STATUS_SUCCESS) {
switch_set_flag(&state, SWITCH_COLLECT_INPUT_DIGITS_DONE);
switch_set_flag(&state, SWITCH_COLLECT_INPUT_SPEECH_DONE);
status = SWITCH_STATUS_SUCCESS;
@@ -3412,9 +3425,9 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_play_and_collect_input(switch_core_se
while ((!switch_test_flag(&state, SWITCH_COLLECT_INPUT_DIGITS_DONE) || !switch_test_flag(&state, SWITCH_COLLECT_INPUT_SPEECH_DONE))
&& switch_channel_ready(channel)) {
status = switch_ivr_sleep(session, sleep_time, SWITCH_FALSE, args);
status = switch_ivr_sleep(session, sleep_time, SWITCH_FALSE, &myargs);
if (args->dmachine && switch_ivr_dmachine_last_ping(args->dmachine) != SWITCH_STATUS_SUCCESS) {
if (args && args->dmachine && switch_ivr_dmachine_last_ping(args->dmachine) != SWITCH_STATUS_SUCCESS) {
// dmachine done
switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_DEBUG, "(%s) DMACHINE DONE\n", switch_channel_get_name(channel));
switch_set_flag(&state, SWITCH_COLLECT_INPUT_DIGITS_DONE);