dect
/
asterisk
Archived
13
0
Fork 0
This repository has been archived on 2022-02-17. You can view files and clone it, but cannot push or open issues or pull requests.
asterisk/res/res_speech.c

354 lines
9.9 KiB
C

/*
* Asterisk -- An open source telephony toolkit.
*
* Copyright (C) 2006, Digium, Inc.
*
* Joshua Colp <jcolp@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 Generic Speech Recognition API
*
* \author Joshua Colp <jcolp@digium.com>
*/
/*** MODULEINFO
<support_level>core</support_level>
***/
#include "asterisk.h"
ASTERISK_FILE_VERSION(__FILE__, "$Revision$");
#include "asterisk/channel.h"
#include "asterisk/module.h"
#include "asterisk/lock.h"
#include "asterisk/linkedlists.h"
#include "asterisk/cli.h"
#include "asterisk/term.h"
#include "asterisk/speech.h"
static AST_RWLIST_HEAD_STATIC(engines, ast_speech_engine);
static struct ast_speech_engine *default_engine = NULL;
/*! \brief Find a speech recognition engine of specified name, if NULL then use the default one */
static struct ast_speech_engine *find_engine(const char *engine_name)
{
struct ast_speech_engine *engine = NULL;
/* If no name is specified -- use the default engine */
if (ast_strlen_zero(engine_name))
return default_engine;
AST_RWLIST_RDLOCK(&engines);
AST_RWLIST_TRAVERSE(&engines, engine, list) {
if (!strcasecmp(engine->name, engine_name)) {
break;
}
}
AST_RWLIST_UNLOCK(&engines);
return engine;
}
/*! \brief Activate a loaded (either local or global) grammar */
int ast_speech_grammar_activate(struct ast_speech *speech, const char *grammar_name)
{
return (speech->engine->activate ? speech->engine->activate(speech, grammar_name) : -1);
}
/*! \brief Deactivate a loaded grammar on a speech structure */
int ast_speech_grammar_deactivate(struct ast_speech *speech, const char *grammar_name)
{
return (speech->engine->deactivate ? speech->engine->deactivate(speech, grammar_name) : -1);
}
/*! \brief Load a local grammar on a speech structure */
int ast_speech_grammar_load(struct ast_speech *speech, const char *grammar_name, const char *grammar)
{
return (speech->engine->load ? speech->engine->load(speech, grammar_name, grammar) : -1);
}
/*! \brief Unload a local grammar from a speech structure */
int ast_speech_grammar_unload(struct ast_speech *speech, const char *grammar_name)
{
return (speech->engine->unload ? speech->engine->unload(speech, grammar_name) : -1);
}
/*! \brief Return the results of a recognition from the speech structure */
struct ast_speech_result *ast_speech_results_get(struct ast_speech *speech)
{
return (speech->engine->get ? speech->engine->get(speech) : NULL);
}
/*! \brief Free a list of results */
int ast_speech_results_free(struct ast_speech_result *result)
{
struct ast_speech_result *current_result = result, *prev_result = NULL;
int res = 0;
while (current_result != NULL) {
prev_result = current_result;
/* Deallocate what we can */
if (current_result->text != NULL) {
ast_free(current_result->text);
current_result->text = NULL;
}
if (current_result->grammar != NULL) {
ast_free(current_result->grammar);
current_result->grammar = NULL;
}
/* Move on and then free ourselves */
current_result = AST_LIST_NEXT(current_result, list);
ast_free(prev_result);
prev_result = NULL;
}
return res;
}
/*! \brief Start speech recognition on a speech structure */
void ast_speech_start(struct ast_speech *speech)
{
/* Clear any flags that may affect things */
ast_clear_flag(speech, AST_SPEECH_SPOKE);
ast_clear_flag(speech, AST_SPEECH_QUIET);
ast_clear_flag(speech, AST_SPEECH_HAVE_RESULTS);
/* If results are on the structure, free them since we are starting again */
if (speech->results) {
ast_speech_results_free(speech->results);
speech->results = NULL;
}
/* If the engine needs to start stuff up, do it */
if (speech->engine->start)
speech->engine->start(speech);
return;
}
/*! \brief Write in signed linear audio to be recognized */
int ast_speech_write(struct ast_speech *speech, void *data, int len)
{
/* Make sure the speech engine is ready to accept audio */
if (speech->state != AST_SPEECH_STATE_READY)
return -1;
return speech->engine->write(speech, data, len);
}
/*! \brief Signal to the engine that DTMF was received */
int ast_speech_dtmf(struct ast_speech *speech, const char *dtmf)
{
int res = 0;
if (speech->state != AST_SPEECH_STATE_READY)
return -1;
if (speech->engine->dtmf != NULL) {
res = speech->engine->dtmf(speech, dtmf);
}
return res;
}
/*! \brief Change an engine specific attribute */
int ast_speech_change(struct ast_speech *speech, const char *name, const char *value)
{
return (speech->engine->change ? speech->engine->change(speech, name, value) : -1);
}
/*! \brief Create a new speech structure using the engine specified */
struct ast_speech *ast_speech_new(const char *engine_name, const struct ast_format_cap *cap)
{
struct ast_speech_engine *engine = NULL;
struct ast_speech *new_speech = NULL;
struct ast_format_cap *joint = NULL;
struct ast_format best;
ast_format_set(&best, AST_FORMAT_SLINEAR, 0);
/* Try to find the speech recognition engine that was requested */
if (!(engine = find_engine(engine_name)))
return NULL;
/* Before even allocating the memory below do some codec negotiation, we choose the best codec possible and fall back to signed linear if possible */
if ((joint = ast_format_cap_joint(engine->formats, cap))) {
ast_best_codec(joint, &best);
joint = ast_format_cap_destroy(joint);
} else if (!ast_format_cap_iscompatible(engine->formats, &best)) {
return NULL;
}
/* Allocate our own speech structure, and try to allocate a structure from the engine too */
if (!(new_speech = ast_calloc(1, sizeof(*new_speech))))
return NULL;
/* Initialize the lock */
ast_mutex_init(&new_speech->lock);
/* Make sure no results are present */
new_speech->results = NULL;
/* Copy over our engine pointer */
new_speech->engine = engine;
/* Can't forget the format audio is going to be in */
ast_format_copy(&new_speech->format, &best);
/* We are not ready to accept audio yet */
ast_speech_change_state(new_speech, AST_SPEECH_STATE_NOT_READY);
/* Pass ourselves to the engine so they can set us up some more and if they error out then do not create a structure */
if (engine->create(new_speech, &best)) {
ast_mutex_destroy(&new_speech->lock);
ast_free(new_speech);
new_speech = NULL;
}
return new_speech;
}
/*! \brief Destroy a speech structure */
int ast_speech_destroy(struct ast_speech *speech)
{
int res = 0;
/* Call our engine so we are destroyed properly */
speech->engine->destroy(speech);
/* Deinitialize the lock */
ast_mutex_destroy(&speech->lock);
/* If results exist on the speech structure, destroy them */
if (speech->results)
ast_speech_results_free(speech->results);
/* If a processing sound is set - free the memory used by it */
if (speech->processing_sound)
ast_free(speech->processing_sound);
/* Aloha we are done */
ast_free(speech);
return res;
}
/*! \brief Change state of a speech structure */
int ast_speech_change_state(struct ast_speech *speech, int state)
{
int res = 0;
switch (state) {
case AST_SPEECH_STATE_WAIT:
/* The engine heard audio, so they spoke */
ast_set_flag(speech, AST_SPEECH_SPOKE);
default:
speech->state = state;
break;
}
return res;
}
/*! \brief Change the type of results we want */
int ast_speech_change_results_type(struct ast_speech *speech, enum ast_speech_results_type results_type)
{
speech->results_type = results_type;
return (speech->engine->change_results_type ? speech->engine->change_results_type(speech, results_type) : 0);
}
/*! \brief Register a speech recognition engine */
int ast_speech_register(struct ast_speech_engine *engine)
{
struct ast_speech_engine *existing_engine = NULL;
int res = 0;
/* Confirm the engine meets the minimum API requirements */
if (!engine->create || !engine->write || !engine->destroy) {
ast_log(LOG_WARNING, "Speech recognition engine '%s' did not meet minimum API requirements.\n", engine->name);
return -1;
}
/* If an engine is already loaded with this name, error out */
if ((existing_engine = find_engine(engine->name))) {
ast_log(LOG_WARNING, "Speech recognition engine '%s' already exists.\n", engine->name);
return -1;
}
ast_verb(2, "Registered speech recognition engine '%s'\n", engine->name);
/* Add to the engine linked list and make default if needed */
AST_RWLIST_WRLOCK(&engines);
AST_RWLIST_INSERT_HEAD(&engines, engine, list);
if (!default_engine) {
default_engine = engine;
ast_verb(2, "Made '%s' the default speech recognition engine\n", engine->name);
}
AST_RWLIST_UNLOCK(&engines);
return res;
}
/*! \brief Unregister a speech recognition engine */
int ast_speech_unregister(const char *engine_name)
{
struct ast_speech_engine *engine = NULL;
int res = -1;
if (ast_strlen_zero(engine_name))
return -1;
AST_RWLIST_WRLOCK(&engines);
AST_RWLIST_TRAVERSE_SAFE_BEGIN(&engines, engine, list) {
if (!strcasecmp(engine->name, engine_name)) {
/* We have our engine... removed it */
AST_RWLIST_REMOVE_CURRENT(list);
/* If this was the default engine, we need to pick a new one */
if (engine == default_engine) {
default_engine = AST_RWLIST_FIRST(&engines);
}
ast_verb(2, "Unregistered speech recognition engine '%s'\n", engine_name);
/* All went well */
res = 0;
break;
}
}
AST_RWLIST_TRAVERSE_SAFE_END;
AST_RWLIST_UNLOCK(&engines);
return res;
}
static int unload_module(void)
{
/* We can not be unloaded */
return -1;
}
static int load_module(void)
{
return AST_MODULE_LOAD_SUCCESS;
}
AST_MODULE_INFO(ASTERISK_GPL_KEY, AST_MODFLAG_GLOBAL_SYMBOLS | AST_MODFLAG_LOAD_ORDER, "Generic Speech Recognition API",
.load = load_module,
.unload = unload_module,
.load_pri = AST_MODPRI_APP_DEPEND,
);