Asterisk - The Open Source Telephony Project GIT-master-3dae2cf
speech.h
Go to the documentation of this file.
1/*
2 * Asterisk -- An open source telephony toolkit.
3 *
4 * Copyright (C) 2006, Digium, Inc.
5 *
6 * Joshua Colp <jcolp@digium.com>
7 *
8 * See http://www.asterisk.org for more information about
9 * the Asterisk project. Please do not directly contact
10 * any of the maintainers of this project for assistance;
11 * the project provides a web site, mailing lists and IRC
12 * channels for your use.
13 *
14 * This program is free software, distributed under the terms of
15 * the GNU General Public License Version 2. See the LICENSE file
16 * at the top of the source tree.
17 */
18
19/*! \file
20 * \brief Generic Speech Recognition API
21 */
22
23#ifndef _ASTERISK_SPEECH_H
24#define _ASTERISK_SPEECH_H
25
26#if defined(__cplusplus) || defined(c_plusplus)
27extern "C" {
28#endif
29
30/* Speech structure flags */
32 AST_SPEECH_QUIET = (1 << 0), /* Quiet down output... they are talking */
33 AST_SPEECH_SPOKE = (1 << 1), /* Speaker spoke! */
34 AST_SPEECH_HAVE_RESULTS = (1 << 2), /* Results are present */
35};
36
37/* Speech structure states - in order of expected change */
39 AST_SPEECH_STATE_NOT_READY = 0, /* Not ready to accept audio */
40 AST_SPEECH_STATE_READY, /* Accepting audio */
41 AST_SPEECH_STATE_WAIT, /* Wait for results to become available */
42 AST_SPEECH_STATE_DONE, /* Processing is all done */
43};
44
48};
49
50/*! \brief Convert a speech results type to a string */
52
53/* Speech structure */
54struct ast_speech {
55 /*! Structure lock */
57 /*! Set flags */
58 unsigned int flags;
59 /*! Processing sound (used when engine is processing audio and getting results) */
61 /*! Current state of structure */
62 int state;
63 /*! Expected write format */
65 /*! Data for speech engine */
66 void *data;
67 /*! Cached results */
69 /*! Type of results we want */
71 /*! Pointer to the engine used by this speech structure */
73};
74
75/* Speech recognition engine structure */
77 /*! Name of speech engine */
78 char *name;
79 /*! Set up the speech structure within the engine */
80 int (*create)(struct ast_speech *speech, struct ast_format *format);
81 /*! Destroy any data set on the speech structure by the engine */
82 int (*destroy)(struct ast_speech *speech);
83 /*! Load a local grammar on the speech structure */
84 int (*load)(struct ast_speech *speech, const char *grammar_name, const char *grammar);
85 /*! Unload a local grammar */
86 int (*unload)(struct ast_speech *speech, const char *grammar_name);
87 /*! Activate a loaded grammar */
88 int (*activate)(struct ast_speech *speech, const char *grammar_name);
89 /*! Deactivate a loaded grammar */
90 int (*deactivate)(struct ast_speech *speech, const char *grammar_name);
91 /*! Write audio to the speech engine */
92 int (*write)(struct ast_speech *speech, void *data, int len);
93 /*! Signal DTMF was received */
94 int (*dtmf)(struct ast_speech *speech, const char *dtmf);
95 /*! Prepare engine to accept audio */
96 int (*start)(struct ast_speech *speech);
97 /*! Change an engine specific setting */
98 int (*change)(struct ast_speech *speech, const char *name, const char *value);
99 /*! Get an engine specific setting */
100 int (*get_setting)(struct ast_speech *speech, const char *name, char *buf, size_t len);
101 /*! Change the type of results we want back */
103 /*! Try to get results */
104 struct ast_speech_result *(*get)(struct ast_speech *speech);
105 /*! Accepted formats by the engine */
108};
109
110/* Result structure */
112 /*! Recognized text */
113 char *text;
114 /*! Result score */
115 int score;
116 /*! NBest Alternative number if in NBest results type */
118 /*! Matched grammar */
119 char *grammar;
120 /*! List information */
122};
123
124/*! \brief Activate a grammar on a speech structure */
125int ast_speech_grammar_activate(struct ast_speech *speech, const char *grammar_name);
126/*! \brief Deactivate a grammar on a speech structure */
127int ast_speech_grammar_deactivate(struct ast_speech *speech, const char *grammar_name);
128/*! \brief Load a grammar on a speech structure (not globally) */
129int ast_speech_grammar_load(struct ast_speech *speech, const char *grammar_name, const char *grammar);
130/*! \brief Unload a grammar */
131int ast_speech_grammar_unload(struct ast_speech *speech, const char *grammar_name);
132/*! \brief Get speech recognition results */
134/*! \brief Free a set of results */
136/*! \brief Indicate to the speech engine that audio is now going to start being written */
137void ast_speech_start(struct ast_speech *speech);
138/*! \brief Create a new speech structure */
139struct ast_speech *ast_speech_new(const char *engine_name, const struct ast_format_cap *formats);
140/*! \brief Destroy a speech structure */
141int ast_speech_destroy(struct ast_speech *speech);
142/*! \brief Write audio to the speech engine */
143int ast_speech_write(struct ast_speech *speech, void *data, int len);
144/*! \brief Signal to the engine that DTMF was received */
145int ast_speech_dtmf(struct ast_speech *speech, const char *dtmf);
146/*! \brief Change an engine specific attribute */
147int ast_speech_change(struct ast_speech *speech, const char *name, const char *value);
148/*! \brief Get an engine specific attribute */
149int ast_speech_get_setting(struct ast_speech *speech, const char *name, char *buf, size_t len);
150/*! \brief Change the type of results we want */
152/*! \brief Change state of a speech structure */
153int ast_speech_change_state(struct ast_speech *speech, int state);
154/*! \brief Register a speech recognition engine */
156/*! \brief Unregister a speech recognition engine */
157int ast_speech_unregister(const char *engine_name);
158/*! \brief Unregister a speech recognition engine */
159struct ast_speech_engine *ast_speech_unregister2(const char *engine_name);
160
161/*! \brief Retrieve a speech recognition engine */
162struct ast_speech_engine *ast_speech_find_engine(const char *engine_name);
163/*! \brief Unregister all speech recognition engines told to by callback */
165 int (*should_unregister)(const struct ast_speech_engine *engine, void *data), void *data,
166 void (*on_unregistered)(void *obj));
167
168#if defined(__cplusplus) || defined(c_plusplus)
169}
170#endif
171
172#endif /* _ASTERISK_SPEECH_H */
static PGresult * result
Definition: cel_pgsql.c:84
static const char type[]
Definition: chan_ooh323.c:109
char buf[BUFSIZE]
Definition: eagi_proxy.c:66
static const char name[]
Definition: format_mp3.c:68
static int len(struct ast_channel *chan, const char *cmd, char *data, char *buf, size_t buflen)
static int should_unregister(const struct ast_speech_engine *engine, void *data)
struct ast_speech_engine * ast_speech_find_engine(const char *engine_name)
Retrieve a speech recognition engine.
Definition: res_speech.c:46
int ast_speech_dtmf(struct ast_speech *speech, const char *dtmf)
Signal to the engine that DTMF was received.
Definition: res_speech.c:154
int ast_speech_results_free(struct ast_speech_result *result)
Free a set of results.
Definition: res_speech.c:96
int ast_speech_grammar_deactivate(struct ast_speech *speech, const char *grammar_name)
Deactivate a grammar on a speech structure.
Definition: res_speech.c:72
const char * ast_speech_results_type_to_string(enum ast_speech_results_type type)
Convert a speech results type to a string.
Definition: res_speech.c:294
int ast_speech_grammar_unload(struct ast_speech *speech, const char *grammar_name)
Unload a grammar.
Definition: res_speech.c:84
void ast_speech_start(struct ast_speech *speech)
Indicate to the speech engine that audio is now going to start being written.
Definition: res_speech.c:122
int ast_speech_unregister(const char *engine_name)
Unregister a speech recognition engine.
Definition: res_speech.c:347
struct ast_speech_engine * ast_speech_unregister2(const char *engine_name)
Unregister a speech recognition engine.
Definition: res_speech.c:352
int ast_speech_write(struct ast_speech *speech, void *data, int len)
Write audio to the speech engine.
Definition: res_speech.c:144
int ast_speech_change(struct ast_speech *speech, const char *name, const char *value)
Change an engine specific attribute.
Definition: res_speech.c:169
struct ast_speech * ast_speech_new(const char *engine_name, const struct ast_format_cap *formats)
Create a new speech structure.
Definition: res_speech.c:181
void ast_speech_unregister_engines(int(*should_unregister)(const struct ast_speech_engine *engine, void *data), void *data, void(*on_unregistered)(void *obj))
Unregister all speech recognition engines told to by callback.
Definition: res_speech.c:380
int ast_speech_destroy(struct ast_speech *speech)
Destroy a speech structure.
Definition: res_speech.c:251
int ast_speech_change_results_type(struct ast_speech *speech, enum ast_speech_results_type results_type)
Change the type of results we want.
Definition: res_speech.c:308
int ast_speech_grammar_load(struct ast_speech *speech, const char *grammar_name, const char *grammar)
Load a grammar on a speech structure (not globally)
Definition: res_speech.c:78
int ast_speech_grammar_activate(struct ast_speech *speech, const char *grammar_name)
Activate a grammar on a speech structure.
Definition: res_speech.c:66
int ast_speech_register(struct ast_speech_engine *engine)
Register a speech recognition engine.
Definition: res_speech.c:316
int ast_speech_get_setting(struct ast_speech *speech, const char *name, char *buf, size_t len)
Get an engine specific attribute.
Definition: res_speech.c:175
int ast_speech_change_state(struct ast_speech *speech, int state)
Change state of a speech structure.
Definition: res_speech.c:278
ast_speech_states
Definition: speech.h:38
@ AST_SPEECH_STATE_DONE
Definition: speech.h:42
@ AST_SPEECH_STATE_READY
Definition: speech.h:40
@ AST_SPEECH_STATE_NOT_READY
Definition: speech.h:39
@ AST_SPEECH_STATE_WAIT
Definition: speech.h:41
ast_speech_results_type
Definition: speech.h:45
@ AST_SPEECH_RESULTS_TYPE_NBEST
Definition: speech.h:47
@ AST_SPEECH_RESULTS_TYPE_NORMAL
Definition: speech.h:46
struct ast_speech_result * ast_speech_results_get(struct ast_speech *speech)
Get speech recognition results.
Definition: res_speech.c:90
ast_speech_flags
Definition: speech.h:31
@ AST_SPEECH_QUIET
Definition: speech.h:32
@ AST_SPEECH_HAVE_RESULTS
Definition: speech.h:34
@ AST_SPEECH_SPOKE
Definition: speech.h:33
Format capabilities structure, holds formats + preference order + etc.
Definition: format_cap.c:54
Definition of a media format.
Definition: format.c:43
Structure for mutex and tracking information.
Definition: lock.h:135
struct ast_format_cap * formats
Definition: speech.h:106
int(* deactivate)(struct ast_speech *speech, const char *grammar_name)
Definition: speech.h:90
int(* start)(struct ast_speech *speech)
Definition: speech.h:96
int(* change_results_type)(struct ast_speech *speech, enum ast_speech_results_type results_type)
Definition: speech.h:102
int(* get_setting)(struct ast_speech *speech, const char *name, char *buf, size_t len)
Definition: speech.h:100
int(* destroy)(struct ast_speech *speech)
Definition: speech.h:82
int(* dtmf)(struct ast_speech *speech, const char *dtmf)
Definition: speech.h:94
int(* load)(struct ast_speech *speech, const char *grammar_name, const char *grammar)
Definition: speech.h:84
int(* change)(struct ast_speech *speech, const char *name, const char *value)
Definition: speech.h:98
int(* activate)(struct ast_speech *speech, const char *grammar_name)
Definition: speech.h:88
char * name
Definition: speech.h:78
int(* create)(struct ast_speech *speech, struct ast_format *format)
Definition: speech.h:80
AST_LIST_ENTRY(ast_speech_engine) list
int(* write)(struct ast_speech *speech, void *data, int len)
Definition: speech.h:92
int(* unload)(struct ast_speech *speech, const char *grammar_name)
Definition: speech.h:86
AST_LIST_ENTRY(ast_speech_result) list
char * grammar
Definition: speech.h:119
char * processing_sound
Definition: speech.h:60
void * data
Definition: speech.h:66
int state
Definition: speech.h:62
struct ast_format * format
Definition: speech.h:64
unsigned int flags
Definition: speech.h:58
struct ast_speech_engine * engine
Definition: speech.h:72
ast_mutex_t lock
Definition: speech.h:56
enum ast_speech_results_type results_type
Definition: speech.h:70
struct ast_speech_result * results
Definition: speech.h:68
Definition: file.c:69
int value
Definition: syslog.c:37