Tristan Matthews | 0a329cc | 2013-07-17 13:20:14 -0400 | [diff] [blame^] | 1 | /* $Id: speex_codec.c 4537 2013-06-19 06:47:43Z riza $ */ |
| 2 | /* |
| 3 | * Copyright (C) 2008-2011 Teluu Inc. (http://www.teluu.com) |
| 4 | * Copyright (C) 2003-2008 Benny Prijono <benny@prijono.org> |
| 5 | * |
| 6 | * This program is free software; you can redistribute it and/or modify |
| 7 | * it under the terms of the GNU General Public License as published by |
| 8 | * the Free Software Foundation; either version 2 of the License, or |
| 9 | * (at your option) any later version. |
| 10 | * |
| 11 | * This program is distributed in the hope that it will be useful, |
| 12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 14 | * GNU General Public License for more details. |
| 15 | * |
| 16 | * You should have received a copy of the GNU General Public License |
| 17 | * along with this program; if not, write to the Free Software |
| 18 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
| 19 | */ |
| 20 | |
| 21 | #include <pjmedia-codec/speex.h> |
| 22 | #include <pjmedia/codec.h> |
| 23 | #include <pjmedia/errno.h> |
| 24 | #include <pjmedia/endpoint.h> |
| 25 | #include <pjmedia/port.h> |
| 26 | #include <speex/speex.h> |
| 27 | #include <pj/assert.h> |
| 28 | #include <pj/log.h> |
| 29 | #include <pj/pool.h> |
| 30 | #include <pj/string.h> |
| 31 | #include <pj/os.h> |
| 32 | |
| 33 | /* |
| 34 | * Only build this file if PJMEDIA_HAS_SPEEX_CODEC != 0 |
| 35 | */ |
| 36 | #if defined(PJMEDIA_HAS_SPEEX_CODEC) && PJMEDIA_HAS_SPEEX_CODEC!=0 |
| 37 | |
| 38 | |
| 39 | #define THIS_FILE "speex_codec.c" |
| 40 | |
| 41 | /* Prototypes for Speex factory */ |
| 42 | static pj_status_t spx_test_alloc( pjmedia_codec_factory *factory, |
| 43 | const pjmedia_codec_info *id ); |
| 44 | static pj_status_t spx_default_attr( pjmedia_codec_factory *factory, |
| 45 | const pjmedia_codec_info *id, |
| 46 | pjmedia_codec_param *attr ); |
| 47 | static pj_status_t spx_enum_codecs( pjmedia_codec_factory *factory, |
| 48 | unsigned *count, |
| 49 | pjmedia_codec_info codecs[]); |
| 50 | static pj_status_t spx_alloc_codec( pjmedia_codec_factory *factory, |
| 51 | const pjmedia_codec_info *id, |
| 52 | pjmedia_codec **p_codec); |
| 53 | static pj_status_t spx_dealloc_codec( pjmedia_codec_factory *factory, |
| 54 | pjmedia_codec *codec ); |
| 55 | |
| 56 | /* Prototypes for Speex implementation. */ |
| 57 | static pj_status_t spx_codec_init( pjmedia_codec *codec, |
| 58 | pj_pool_t *pool ); |
| 59 | static pj_status_t spx_codec_open( pjmedia_codec *codec, |
| 60 | pjmedia_codec_param *attr ); |
| 61 | static pj_status_t spx_codec_close( pjmedia_codec *codec ); |
| 62 | static pj_status_t spx_codec_modify(pjmedia_codec *codec, |
| 63 | const pjmedia_codec_param *attr ); |
| 64 | static pj_status_t spx_codec_parse( pjmedia_codec *codec, |
| 65 | void *pkt, |
| 66 | pj_size_t pkt_size, |
| 67 | const pj_timestamp *ts, |
| 68 | unsigned *frame_cnt, |
| 69 | pjmedia_frame frames[]); |
| 70 | static pj_status_t spx_codec_encode( pjmedia_codec *codec, |
| 71 | const struct pjmedia_frame *input, |
| 72 | unsigned output_buf_len, |
| 73 | struct pjmedia_frame *output); |
| 74 | static pj_status_t spx_codec_decode( pjmedia_codec *codec, |
| 75 | const struct pjmedia_frame *input, |
| 76 | unsigned output_buf_len, |
| 77 | struct pjmedia_frame *output); |
| 78 | static pj_status_t spx_codec_recover(pjmedia_codec *codec, |
| 79 | unsigned output_buf_len, |
| 80 | struct pjmedia_frame *output); |
| 81 | |
| 82 | /* Definition for Speex codec operations. */ |
| 83 | static pjmedia_codec_op spx_op = |
| 84 | { |
| 85 | &spx_codec_init, |
| 86 | &spx_codec_open, |
| 87 | &spx_codec_close, |
| 88 | &spx_codec_modify, |
| 89 | &spx_codec_parse, |
| 90 | &spx_codec_encode, |
| 91 | &spx_codec_decode, |
| 92 | &spx_codec_recover |
| 93 | }; |
| 94 | |
| 95 | /* Definition for Speex codec factory operations. */ |
| 96 | static pjmedia_codec_factory_op spx_factory_op = |
| 97 | { |
| 98 | &spx_test_alloc, |
| 99 | &spx_default_attr, |
| 100 | &spx_enum_codecs, |
| 101 | &spx_alloc_codec, |
| 102 | &spx_dealloc_codec, |
| 103 | &pjmedia_codec_speex_deinit |
| 104 | }; |
| 105 | |
| 106 | /* Index to Speex parameter. */ |
| 107 | enum |
| 108 | { |
| 109 | PARAM_NB, /* Index for narrowband parameter. */ |
| 110 | PARAM_WB, /* Index for wideband parameter. */ |
| 111 | PARAM_UWB, /* Index for ultra-wideband parameter */ |
| 112 | }; |
| 113 | |
| 114 | /* Speex default parameter */ |
| 115 | struct speex_param |
| 116 | { |
| 117 | int enabled; /* Is this mode enabled? */ |
| 118 | const SpeexMode *mode; /* Speex mode. */ |
| 119 | int pt; /* Payload type. */ |
| 120 | unsigned clock_rate; /* Default sampling rate to be used.*/ |
| 121 | int quality; /* Default encoder quality. */ |
| 122 | int complexity; /* Default encoder complexity. */ |
| 123 | int samples_per_frame; /* Samples per frame. */ |
| 124 | int framesize; /* Frame size for current mode. */ |
| 125 | int bitrate; /* Bit rate for current mode. */ |
| 126 | int max_bitrate; /* Max bit rate for current mode. */ |
| 127 | }; |
| 128 | |
| 129 | /* Speex factory */ |
| 130 | static struct spx_factory |
| 131 | { |
| 132 | pjmedia_codec_factory base; |
| 133 | pjmedia_endpt *endpt; |
| 134 | pj_pool_t *pool; |
| 135 | pj_mutex_t *mutex; |
| 136 | pjmedia_codec codec_list; |
| 137 | struct speex_param speex_param[3]; |
| 138 | |
| 139 | } spx_factory; |
| 140 | |
| 141 | /* Speex codec private data. */ |
| 142 | struct spx_private |
| 143 | { |
| 144 | int param_id; /**< Index to speex param. */ |
| 145 | |
| 146 | void *enc; /**< Encoder state. */ |
| 147 | SpeexBits enc_bits; /**< Encoder bits. */ |
| 148 | void *dec; /**< Decoder state. */ |
| 149 | SpeexBits dec_bits; /**< Decoder bits. */ |
| 150 | }; |
| 151 | |
| 152 | |
| 153 | /* |
| 154 | * Get codec bitrate and frame size. |
| 155 | */ |
| 156 | static pj_status_t get_speex_info( struct speex_param *p ) |
| 157 | { |
| 158 | void *state; |
| 159 | int tmp; |
| 160 | |
| 161 | /* Create temporary encoder */ |
| 162 | state = speex_encoder_init(p->mode); |
| 163 | if (!state) |
| 164 | return PJMEDIA_CODEC_EFAILED; |
| 165 | |
| 166 | /* Set the quality */ |
| 167 | if (p->quality != -1) |
| 168 | speex_encoder_ctl(state, SPEEX_SET_QUALITY, &p->quality); |
| 169 | |
| 170 | /* Sampling rate. */ |
| 171 | speex_encoder_ctl(state, SPEEX_SET_SAMPLING_RATE, &p->clock_rate); |
| 172 | |
| 173 | /* VAD off to have max bitrate */ |
| 174 | tmp = 0; |
| 175 | speex_encoder_ctl(state, SPEEX_SET_VAD, &tmp); |
| 176 | |
| 177 | /* Complexity. */ |
| 178 | if (p->complexity != -1) |
| 179 | speex_encoder_ctl(state, SPEEX_SET_COMPLEXITY, &p->complexity); |
| 180 | |
| 181 | /* Now get the frame size */ |
| 182 | speex_encoder_ctl(state, SPEEX_GET_FRAME_SIZE, &p->samples_per_frame); |
| 183 | |
| 184 | /* Now get the average bitrate */ |
| 185 | speex_encoder_ctl(state, SPEEX_GET_BITRATE, &p->bitrate); |
| 186 | |
| 187 | /* Calculate framesize. */ |
| 188 | p->framesize = p->bitrate * 20 / 1000; |
| 189 | |
| 190 | /* Now get the maximum bitrate by using maximum quality (=10) */ |
| 191 | tmp = 10; |
| 192 | speex_encoder_ctl(state, SPEEX_SET_QUALITY, &tmp); |
| 193 | speex_encoder_ctl(state, SPEEX_GET_BITRATE, &p->max_bitrate); |
| 194 | |
| 195 | /* Destroy encoder. */ |
| 196 | speex_encoder_destroy(state); |
| 197 | |
| 198 | return PJ_SUCCESS; |
| 199 | } |
| 200 | |
| 201 | /* |
| 202 | * Initialize and register Speex codec factory to pjmedia endpoint. |
| 203 | */ |
| 204 | PJ_DEF(pj_status_t) pjmedia_codec_speex_init( pjmedia_endpt *endpt, |
| 205 | unsigned options, |
| 206 | int quality, |
| 207 | int complexity ) |
| 208 | { |
| 209 | pjmedia_codec_mgr *codec_mgr; |
| 210 | unsigned i; |
| 211 | pj_status_t status; |
| 212 | |
| 213 | if (spx_factory.pool != NULL) { |
| 214 | /* Already initialized. */ |
| 215 | return PJ_SUCCESS; |
| 216 | } |
| 217 | |
| 218 | /* Get defaults */ |
| 219 | if (quality < 0) quality = PJMEDIA_CODEC_SPEEX_DEFAULT_QUALITY; |
| 220 | if (complexity < 0) complexity = PJMEDIA_CODEC_SPEEX_DEFAULT_COMPLEXITY; |
| 221 | |
| 222 | /* Validate quality & complexity */ |
| 223 | PJ_ASSERT_RETURN(quality >= 0 && quality <= 10, PJ_EINVAL); |
| 224 | PJ_ASSERT_RETURN(complexity >= 1 && complexity <= 10, PJ_EINVAL); |
| 225 | |
| 226 | /* Create Speex codec factory. */ |
| 227 | spx_factory.base.op = &spx_factory_op; |
| 228 | spx_factory.base.factory_data = NULL; |
| 229 | spx_factory.endpt = endpt; |
| 230 | |
| 231 | spx_factory.pool = pjmedia_endpt_create_pool(endpt, "speex", |
| 232 | 4000, 4000); |
| 233 | if (!spx_factory.pool) |
| 234 | return PJ_ENOMEM; |
| 235 | |
| 236 | pj_list_init(&spx_factory.codec_list); |
| 237 | |
| 238 | /* Create mutex. */ |
| 239 | status = pj_mutex_create_simple(spx_factory.pool, "speex", |
| 240 | &spx_factory.mutex); |
| 241 | if (status != PJ_SUCCESS) |
| 242 | goto on_error; |
| 243 | |
| 244 | /* Initialize default Speex parameter. */ |
| 245 | spx_factory.speex_param[PARAM_NB].enabled = |
| 246 | ((options & PJMEDIA_SPEEX_NO_NB) == 0); |
| 247 | spx_factory.speex_param[PARAM_NB].pt = PJMEDIA_RTP_PT_SPEEX_NB; |
| 248 | spx_factory.speex_param[PARAM_NB].mode = speex_lib_get_mode(SPEEX_MODEID_NB); |
| 249 | spx_factory.speex_param[PARAM_NB].clock_rate = 8000; |
| 250 | spx_factory.speex_param[PARAM_NB].quality = quality; |
| 251 | spx_factory.speex_param[PARAM_NB].complexity = complexity; |
| 252 | |
| 253 | spx_factory.speex_param[PARAM_WB].enabled = |
| 254 | ((options & PJMEDIA_SPEEX_NO_WB) == 0); |
| 255 | spx_factory.speex_param[PARAM_WB].pt = PJMEDIA_RTP_PT_SPEEX_WB; |
| 256 | spx_factory.speex_param[PARAM_WB].mode = speex_lib_get_mode(SPEEX_MODEID_WB); |
| 257 | spx_factory.speex_param[PARAM_WB].clock_rate = 16000; |
| 258 | spx_factory.speex_param[PARAM_WB].quality = quality; |
| 259 | spx_factory.speex_param[PARAM_WB].complexity = complexity; |
| 260 | |
| 261 | spx_factory.speex_param[PARAM_UWB].enabled = |
| 262 | ((options & PJMEDIA_SPEEX_NO_UWB) == 0); |
| 263 | spx_factory.speex_param[PARAM_UWB].pt = PJMEDIA_RTP_PT_SPEEX_UWB; |
| 264 | spx_factory.speex_param[PARAM_UWB].mode = speex_lib_get_mode(SPEEX_MODEID_UWB); |
| 265 | spx_factory.speex_param[PARAM_UWB].clock_rate = 32000; |
| 266 | spx_factory.speex_param[PARAM_UWB].quality = quality; |
| 267 | spx_factory.speex_param[PARAM_UWB].complexity = complexity; |
| 268 | |
| 269 | /* Somehow quality <=4 is broken in linux. */ |
| 270 | if (quality <= 4 && quality >= 0) { |
| 271 | PJ_LOG(5,(THIS_FILE, "Adjusting quality to 5 for uwb")); |
| 272 | spx_factory.speex_param[PARAM_UWB].quality = 5; |
| 273 | } |
| 274 | |
| 275 | /* Get codec framesize and avg bitrate for each mode. */ |
| 276 | for (i=0; i<PJ_ARRAY_SIZE(spx_factory.speex_param); ++i) { |
| 277 | status = get_speex_info(&spx_factory.speex_param[i]); |
| 278 | } |
| 279 | |
| 280 | /* Get the codec manager. */ |
| 281 | codec_mgr = pjmedia_endpt_get_codec_mgr(endpt); |
| 282 | if (!codec_mgr) { |
| 283 | status = PJ_EINVALIDOP; |
| 284 | goto on_error; |
| 285 | } |
| 286 | |
| 287 | /* Register codec factory to endpoint. */ |
| 288 | status = pjmedia_codec_mgr_register_factory(codec_mgr, |
| 289 | &spx_factory.base); |
| 290 | if (status != PJ_SUCCESS) |
| 291 | goto on_error; |
| 292 | |
| 293 | /* Done. */ |
| 294 | return PJ_SUCCESS; |
| 295 | |
| 296 | on_error: |
| 297 | pj_pool_release(spx_factory.pool); |
| 298 | spx_factory.pool = NULL; |
| 299 | return status; |
| 300 | } |
| 301 | |
| 302 | |
| 303 | /* |
| 304 | * Initialize with default settings. |
| 305 | */ |
| 306 | PJ_DEF(pj_status_t) pjmedia_codec_speex_init_default(pjmedia_endpt *endpt) |
| 307 | { |
| 308 | return pjmedia_codec_speex_init(endpt, 0, -1, -1); |
| 309 | } |
| 310 | |
| 311 | /* |
| 312 | * Change the settings of Speex codec. |
| 313 | */ |
| 314 | PJ_DEF(pj_status_t) pjmedia_codec_speex_set_param(unsigned clock_rate, |
| 315 | int quality, |
| 316 | int complexity) |
| 317 | { |
| 318 | unsigned i; |
| 319 | |
| 320 | /* Get defaults */ |
| 321 | if (quality < 0) quality = PJMEDIA_CODEC_SPEEX_DEFAULT_QUALITY; |
| 322 | if (complexity < 0) complexity = PJMEDIA_CODEC_SPEEX_DEFAULT_COMPLEXITY; |
| 323 | |
| 324 | /* Validate quality & complexity */ |
| 325 | PJ_ASSERT_RETURN(quality >= 0 && quality <= 10, PJ_EINVAL); |
| 326 | PJ_ASSERT_RETURN(complexity >= 1 && complexity <= 10, PJ_EINVAL); |
| 327 | |
| 328 | /* Apply the settings */ |
| 329 | for (i=0; i<PJ_ARRAY_SIZE(spx_factory.speex_param); ++i) { |
| 330 | if (spx_factory.speex_param[i].clock_rate == clock_rate) { |
| 331 | pj_status_t status; |
| 332 | |
| 333 | spx_factory.speex_param[i].quality = quality; |
| 334 | spx_factory.speex_param[i].complexity = complexity; |
| 335 | |
| 336 | /* Somehow quality<=4 is broken in linux. */ |
| 337 | if (i == PARAM_UWB && quality <= 4 && quality >= 0) { |
| 338 | PJ_LOG(5,(THIS_FILE, "Adjusting quality to 5 for uwb")); |
| 339 | spx_factory.speex_param[PARAM_UWB].quality = 5; |
| 340 | } |
| 341 | |
| 342 | status = get_speex_info(&spx_factory.speex_param[i]); |
| 343 | |
| 344 | return status; |
| 345 | } |
| 346 | } |
| 347 | |
| 348 | return PJ_EINVAL; |
| 349 | } |
| 350 | |
| 351 | /* |
| 352 | * Unregister Speex codec factory from pjmedia endpoint and deinitialize |
| 353 | * the Speex codec library. |
| 354 | */ |
| 355 | PJ_DEF(pj_status_t) pjmedia_codec_speex_deinit(void) |
| 356 | { |
| 357 | pjmedia_codec_mgr *codec_mgr; |
| 358 | pj_status_t status; |
| 359 | |
| 360 | if (spx_factory.pool == NULL) { |
| 361 | /* Already deinitialized */ |
| 362 | return PJ_SUCCESS; |
| 363 | } |
| 364 | |
| 365 | pj_mutex_lock(spx_factory.mutex); |
| 366 | |
| 367 | /* We don't want to deinit if there's outstanding codec. */ |
| 368 | /* This is silly, as we'll always have codec in the list if |
| 369 | we ever allocate a codec! A better behavior maybe is to |
| 370 | deallocate all codecs in the list. |
| 371 | if (!pj_list_empty(&spx_factory.codec_list)) { |
| 372 | pj_mutex_unlock(spx_factory.mutex); |
| 373 | return PJ_EBUSY; |
| 374 | } |
| 375 | */ |
| 376 | |
| 377 | /* Get the codec manager. */ |
| 378 | codec_mgr = pjmedia_endpt_get_codec_mgr(spx_factory.endpt); |
| 379 | if (!codec_mgr) { |
| 380 | pj_pool_release(spx_factory.pool); |
| 381 | spx_factory.pool = NULL; |
| 382 | return PJ_EINVALIDOP; |
| 383 | } |
| 384 | |
| 385 | /* Unregister Speex codec factory. */ |
| 386 | status = pjmedia_codec_mgr_unregister_factory(codec_mgr, |
| 387 | &spx_factory.base); |
| 388 | |
| 389 | /* Destroy mutex. */ |
| 390 | pj_mutex_destroy(spx_factory.mutex); |
| 391 | |
| 392 | /* Destroy pool. */ |
| 393 | pj_pool_release(spx_factory.pool); |
| 394 | spx_factory.pool = NULL; |
| 395 | |
| 396 | return status; |
| 397 | } |
| 398 | |
| 399 | /* |
| 400 | * Check if factory can allocate the specified codec. |
| 401 | */ |
| 402 | static pj_status_t spx_test_alloc( pjmedia_codec_factory *factory, |
| 403 | const pjmedia_codec_info *info ) |
| 404 | { |
| 405 | const pj_str_t speex_tag = { "speex", 5}; |
| 406 | unsigned i; |
| 407 | |
| 408 | PJ_UNUSED_ARG(factory); |
| 409 | |
| 410 | /* Type MUST be audio. */ |
| 411 | if (info->type != PJMEDIA_TYPE_AUDIO) |
| 412 | return PJMEDIA_CODEC_EUNSUP; |
| 413 | |
| 414 | /* Check encoding name. */ |
| 415 | if (pj_stricmp(&info->encoding_name, &speex_tag) != 0) |
| 416 | return PJMEDIA_CODEC_EUNSUP; |
| 417 | |
| 418 | /* Check clock-rate */ |
| 419 | for (i=0; i<PJ_ARRAY_SIZE(spx_factory.speex_param); ++i) { |
| 420 | if (info->clock_rate == spx_factory.speex_param[i].clock_rate) { |
| 421 | /* Okay, let's Speex! */ |
| 422 | return PJ_SUCCESS; |
| 423 | } |
| 424 | } |
| 425 | |
| 426 | |
| 427 | /* Unsupported, or mode is disabled. */ |
| 428 | return PJMEDIA_CODEC_EUNSUP; |
| 429 | } |
| 430 | |
| 431 | /* |
| 432 | * Generate default attribute. |
| 433 | */ |
| 434 | static pj_status_t spx_default_attr (pjmedia_codec_factory *factory, |
| 435 | const pjmedia_codec_info *id, |
| 436 | pjmedia_codec_param *attr ) |
| 437 | { |
| 438 | |
| 439 | PJ_ASSERT_RETURN(factory==&spx_factory.base, PJ_EINVAL); |
| 440 | |
| 441 | pj_bzero(attr, sizeof(pjmedia_codec_param)); |
| 442 | attr->info.pt = (pj_uint8_t)id->pt; |
| 443 | attr->info.channel_cnt = 1; |
| 444 | |
| 445 | if (id->clock_rate <= 8000) { |
| 446 | attr->info.clock_rate = spx_factory.speex_param[PARAM_NB].clock_rate; |
| 447 | attr->info.avg_bps = spx_factory.speex_param[PARAM_NB].bitrate; |
| 448 | attr->info.max_bps = spx_factory.speex_param[PARAM_NB].max_bitrate; |
| 449 | |
| 450 | } else if (id->clock_rate <= 16000) { |
| 451 | attr->info.clock_rate = spx_factory.speex_param[PARAM_WB].clock_rate; |
| 452 | attr->info.avg_bps = spx_factory.speex_param[PARAM_WB].bitrate; |
| 453 | attr->info.max_bps = spx_factory.speex_param[PARAM_WB].max_bitrate; |
| 454 | |
| 455 | } else { |
| 456 | /* Wow.. somebody is doing ultra-wideband. Cool...! */ |
| 457 | attr->info.clock_rate = spx_factory.speex_param[PARAM_UWB].clock_rate; |
| 458 | attr->info.avg_bps = spx_factory.speex_param[PARAM_UWB].bitrate; |
| 459 | attr->info.max_bps = spx_factory.speex_param[PARAM_UWB].max_bitrate; |
| 460 | } |
| 461 | |
| 462 | attr->info.pcm_bits_per_sample = 16; |
| 463 | attr->info.frm_ptime = 20; |
| 464 | attr->info.pt = (pj_uint8_t)id->pt; |
| 465 | |
| 466 | attr->setting.frm_per_pkt = 1; |
| 467 | |
| 468 | /* Default flags. */ |
| 469 | attr->setting.cng = 1; |
| 470 | attr->setting.plc = 1; |
| 471 | attr->setting.penh =1 ; |
| 472 | attr->setting.vad = 1; |
| 473 | |
| 474 | return PJ_SUCCESS; |
| 475 | } |
| 476 | |
| 477 | /* |
| 478 | * Enum codecs supported by this factory (i.e. only Speex!). |
| 479 | */ |
| 480 | static pj_status_t spx_enum_codecs(pjmedia_codec_factory *factory, |
| 481 | unsigned *count, |
| 482 | pjmedia_codec_info codecs[]) |
| 483 | { |
| 484 | unsigned max; |
| 485 | int i; /* Must be signed */ |
| 486 | |
| 487 | PJ_UNUSED_ARG(factory); |
| 488 | PJ_ASSERT_RETURN(codecs && *count > 0, PJ_EINVAL); |
| 489 | |
| 490 | max = *count; |
| 491 | *count = 0; |
| 492 | |
| 493 | /* |
| 494 | * We return three codecs here, and in this order: |
| 495 | * - ultra-wideband, wideband, and narrowband. |
| 496 | */ |
| 497 | for (i=PJ_ARRAY_SIZE(spx_factory.speex_param)-1; i>=0 && *count<max; --i) { |
| 498 | |
| 499 | if (!spx_factory.speex_param[i].enabled) |
| 500 | continue; |
| 501 | |
| 502 | pj_bzero(&codecs[*count], sizeof(pjmedia_codec_info)); |
| 503 | codecs[*count].encoding_name = pj_str("speex"); |
| 504 | codecs[*count].pt = spx_factory.speex_param[i].pt; |
| 505 | codecs[*count].type = PJMEDIA_TYPE_AUDIO; |
| 506 | codecs[*count].clock_rate = spx_factory.speex_param[i].clock_rate; |
| 507 | codecs[*count].channel_cnt = 1; |
| 508 | |
| 509 | ++*count; |
| 510 | } |
| 511 | |
| 512 | return PJ_SUCCESS; |
| 513 | } |
| 514 | |
| 515 | /* |
| 516 | * Allocate a new Speex codec instance. |
| 517 | */ |
| 518 | static pj_status_t spx_alloc_codec( pjmedia_codec_factory *factory, |
| 519 | const pjmedia_codec_info *id, |
| 520 | pjmedia_codec **p_codec) |
| 521 | { |
| 522 | pjmedia_codec *codec; |
| 523 | struct spx_private *spx; |
| 524 | |
| 525 | PJ_ASSERT_RETURN(factory && id && p_codec, PJ_EINVAL); |
| 526 | PJ_ASSERT_RETURN(factory == &spx_factory.base, PJ_EINVAL); |
| 527 | |
| 528 | |
| 529 | pj_mutex_lock(spx_factory.mutex); |
| 530 | |
| 531 | /* Get free nodes, if any. */ |
| 532 | if (!pj_list_empty(&spx_factory.codec_list)) { |
| 533 | codec = spx_factory.codec_list.next; |
| 534 | pj_list_erase(codec); |
| 535 | } else { |
| 536 | codec = PJ_POOL_ZALLOC_T(spx_factory.pool, pjmedia_codec); |
| 537 | PJ_ASSERT_RETURN(codec != NULL, PJ_ENOMEM); |
| 538 | codec->op = &spx_op; |
| 539 | codec->factory = factory; |
| 540 | codec->codec_data = pj_pool_alloc(spx_factory.pool, |
| 541 | sizeof(struct spx_private)); |
| 542 | } |
| 543 | |
| 544 | pj_mutex_unlock(spx_factory.mutex); |
| 545 | |
| 546 | spx = (struct spx_private*) codec->codec_data; |
| 547 | spx->enc = NULL; |
| 548 | spx->dec = NULL; |
| 549 | |
| 550 | if (id->clock_rate <= 8000) |
| 551 | spx->param_id = PARAM_NB; |
| 552 | else if (id->clock_rate <= 16000) |
| 553 | spx->param_id = PARAM_WB; |
| 554 | else |
| 555 | spx->param_id = PARAM_UWB; |
| 556 | |
| 557 | *p_codec = codec; |
| 558 | return PJ_SUCCESS; |
| 559 | } |
| 560 | |
| 561 | /* |
| 562 | * Free codec. |
| 563 | */ |
| 564 | static pj_status_t spx_dealloc_codec( pjmedia_codec_factory *factory, |
| 565 | pjmedia_codec *codec ) |
| 566 | { |
| 567 | struct spx_private *spx; |
| 568 | |
| 569 | PJ_ASSERT_RETURN(factory && codec, PJ_EINVAL); |
| 570 | PJ_ASSERT_RETURN(factory == &spx_factory.base, PJ_EINVAL); |
| 571 | |
| 572 | /* Close codec, if it's not closed. */ |
| 573 | spx = (struct spx_private*) codec->codec_data; |
| 574 | if (spx->enc != NULL || spx->dec != NULL) { |
| 575 | spx_codec_close(codec); |
| 576 | } |
| 577 | |
| 578 | /* Put in the free list. */ |
| 579 | pj_mutex_lock(spx_factory.mutex); |
| 580 | pj_list_push_front(&spx_factory.codec_list, codec); |
| 581 | pj_mutex_unlock(spx_factory.mutex); |
| 582 | |
| 583 | return PJ_SUCCESS; |
| 584 | } |
| 585 | |
| 586 | /* |
| 587 | * Init codec. |
| 588 | */ |
| 589 | static pj_status_t spx_codec_init( pjmedia_codec *codec, |
| 590 | pj_pool_t *pool ) |
| 591 | { |
| 592 | PJ_UNUSED_ARG(codec); |
| 593 | PJ_UNUSED_ARG(pool); |
| 594 | return PJ_SUCCESS; |
| 595 | } |
| 596 | |
| 597 | /* |
| 598 | * Open codec. |
| 599 | */ |
| 600 | static pj_status_t spx_codec_open( pjmedia_codec *codec, |
| 601 | pjmedia_codec_param *attr ) |
| 602 | { |
| 603 | struct spx_private *spx; |
| 604 | int id, tmp; |
| 605 | |
| 606 | spx = (struct spx_private*) codec->codec_data; |
| 607 | id = spx->param_id; |
| 608 | |
| 609 | /* |
| 610 | * Create and initialize encoder. |
| 611 | */ |
| 612 | spx->enc = speex_encoder_init(spx_factory.speex_param[id].mode); |
| 613 | if (!spx->enc) |
| 614 | return PJMEDIA_CODEC_EFAILED; |
| 615 | speex_bits_init(&spx->enc_bits); |
| 616 | |
| 617 | /* Set the quality*/ |
| 618 | if (spx_factory.speex_param[id].quality != -1) { |
| 619 | speex_encoder_ctl(spx->enc, SPEEX_SET_QUALITY, |
| 620 | &spx_factory.speex_param[id].quality); |
| 621 | } |
| 622 | |
| 623 | /* Sampling rate. */ |
| 624 | tmp = attr->info.clock_rate; |
| 625 | speex_encoder_ctl(spx->enc, SPEEX_SET_SAMPLING_RATE, |
| 626 | &spx_factory.speex_param[id].clock_rate); |
| 627 | |
| 628 | /* VAD */ |
| 629 | tmp = (attr->setting.vad != 0); |
| 630 | speex_encoder_ctl(spx->enc, SPEEX_SET_VAD, &tmp); |
| 631 | speex_encoder_ctl(spx->enc, SPEEX_SET_DTX, &tmp); |
| 632 | |
| 633 | /* Complexity */ |
| 634 | if (spx_factory.speex_param[id].complexity != -1) { |
| 635 | speex_encoder_ctl(spx->enc, SPEEX_SET_COMPLEXITY, |
| 636 | &spx_factory.speex_param[id].complexity); |
| 637 | } |
| 638 | |
| 639 | /* |
| 640 | * Create and initialize decoder. |
| 641 | */ |
| 642 | spx->dec = speex_decoder_init(spx_factory.speex_param[id].mode); |
| 643 | if (!spx->dec) { |
| 644 | spx_codec_close(codec); |
| 645 | return PJMEDIA_CODEC_EFAILED; |
| 646 | } |
| 647 | speex_bits_init(&spx->dec_bits); |
| 648 | |
| 649 | /* Sampling rate. */ |
| 650 | speex_decoder_ctl(spx->dec, SPEEX_SET_SAMPLING_RATE, |
| 651 | &spx_factory.speex_param[id].clock_rate); |
| 652 | |
| 653 | /* PENH */ |
| 654 | tmp = attr->setting.penh; |
| 655 | speex_decoder_ctl(spx->dec, SPEEX_SET_ENH, &tmp); |
| 656 | |
| 657 | return PJ_SUCCESS; |
| 658 | } |
| 659 | |
| 660 | /* |
| 661 | * Close codec. |
| 662 | */ |
| 663 | static pj_status_t spx_codec_close( pjmedia_codec *codec ) |
| 664 | { |
| 665 | struct spx_private *spx; |
| 666 | |
| 667 | spx = (struct spx_private*) codec->codec_data; |
| 668 | |
| 669 | /* Destroy encoder*/ |
| 670 | if (spx->enc) { |
| 671 | speex_encoder_destroy( spx->enc ); |
| 672 | spx->enc = NULL; |
| 673 | speex_bits_destroy( &spx->enc_bits ); |
| 674 | } |
| 675 | |
| 676 | /* Destroy decoder */ |
| 677 | if (spx->dec) { |
| 678 | speex_decoder_destroy( spx->dec); |
| 679 | spx->dec = NULL; |
| 680 | speex_bits_destroy( &spx->dec_bits ); |
| 681 | } |
| 682 | |
| 683 | return PJ_SUCCESS; |
| 684 | } |
| 685 | |
| 686 | |
| 687 | /* |
| 688 | * Modify codec settings. |
| 689 | */ |
| 690 | static pj_status_t spx_codec_modify(pjmedia_codec *codec, |
| 691 | const pjmedia_codec_param *attr ) |
| 692 | { |
| 693 | struct spx_private *spx; |
| 694 | int tmp; |
| 695 | |
| 696 | spx = (struct spx_private*) codec->codec_data; |
| 697 | |
| 698 | /* VAD */ |
| 699 | tmp = (attr->setting.vad != 0); |
| 700 | speex_encoder_ctl(spx->enc, SPEEX_SET_VAD, &tmp); |
| 701 | speex_encoder_ctl(spx->enc, SPEEX_SET_DTX, &tmp); |
| 702 | |
| 703 | /* PENH */ |
| 704 | tmp = attr->setting.penh; |
| 705 | speex_decoder_ctl(spx->dec, SPEEX_SET_ENH, &tmp); |
| 706 | |
| 707 | return PJ_SUCCESS; |
| 708 | } |
| 709 | |
| 710 | #if 0 |
| 711 | # define TRACE__(args) PJ_LOG(5,args) |
| 712 | #else |
| 713 | # define TRACE__(args) |
| 714 | #endif |
| 715 | |
| 716 | #undef THIS_FUNC |
| 717 | #define THIS_FUNC "speex_get_next_frame" |
| 718 | |
| 719 | #define NB_SUBMODES 16 |
| 720 | #define NB_SUBMODE_BITS 4 |
| 721 | |
| 722 | #define SB_SUBMODES 8 |
| 723 | #define SB_SUBMODE_BITS 3 |
| 724 | |
| 725 | /* This function will iterate frames & submodes in the Speex bits. |
| 726 | * Returns 0 if a frame found, otherwise returns -1. |
| 727 | */ |
| 728 | int speex_get_next_frame(SpeexBits *bits) |
| 729 | { |
| 730 | static const int inband_skip_table[NB_SUBMODES] = |
| 731 | {1, 1, 4, 4, 4, 4, 4, 4, 8, 8, 16, 16, 32, 32, 64, 64 }; |
| 732 | static const int wb_skip_table[SB_SUBMODES] = |
| 733 | {SB_SUBMODE_BITS+1, 36, 112, 192, 352, -1, -1, -1}; |
| 734 | |
| 735 | unsigned submode; |
| 736 | unsigned nb_count = 0; |
| 737 | |
| 738 | while (speex_bits_remaining(bits) >= 5) { |
| 739 | unsigned wb_count = 0; |
| 740 | unsigned bit_ptr = bits->bitPtr; |
| 741 | unsigned char_ptr = bits->charPtr; |
| 742 | |
| 743 | /* WB frame */ |
| 744 | while ((speex_bits_remaining(bits) >= 4) |
| 745 | && speex_bits_unpack_unsigned(bits, 1)) |
| 746 | { |
| 747 | int advance; |
| 748 | |
| 749 | submode = speex_bits_unpack_unsigned(bits, 3); |
| 750 | advance = wb_skip_table[submode]; |
| 751 | if (advance < 0) { |
| 752 | TRACE__((THIS_FUNC, "Invalid mode encountered. " |
| 753 | "The stream is corrupted.")); |
| 754 | return -1; |
| 755 | } |
| 756 | TRACE__((THIS_FUNC, "WB layer skipped: %d bits", advance)); |
| 757 | advance -= (SB_SUBMODE_BITS+1); |
| 758 | speex_bits_advance(bits, advance); |
| 759 | |
| 760 | bit_ptr = bits->bitPtr; |
| 761 | char_ptr = bits->charPtr; |
| 762 | |
| 763 | /* Consecutive subband frames may not exceed 2 frames */ |
| 764 | if (++wb_count > 2) |
| 765 | return -1; |
| 766 | } |
| 767 | |
| 768 | /* End of bits, return the frame */ |
| 769 | if (speex_bits_remaining(bits) < 4) { |
| 770 | TRACE__((THIS_FUNC, "End of stream")); |
| 771 | return 0; |
| 772 | } |
| 773 | |
| 774 | /* Stop iteration, return the frame */ |
| 775 | if (nb_count > 0) { |
| 776 | bits->bitPtr = bit_ptr; |
| 777 | bits->charPtr = char_ptr; |
| 778 | return 0; |
| 779 | } |
| 780 | |
| 781 | /* Get control bits */ |
| 782 | submode = speex_bits_unpack_unsigned(bits, 4); |
| 783 | TRACE__((THIS_FUNC, "Control bits: %d at %d", |
| 784 | submode, bits->charPtr*8+bits->bitPtr)); |
| 785 | |
| 786 | if (submode == 15) { |
| 787 | TRACE__((THIS_FUNC, "Found submode: terminator")); |
| 788 | return -1; |
| 789 | } else if (submode == 14) { |
| 790 | /* in-band signal; next 4 bits contain signal id */ |
| 791 | submode = speex_bits_unpack_unsigned(bits, 4); |
| 792 | TRACE__((THIS_FUNC, "Found submode: in-band %d bits", |
| 793 | inband_skip_table[submode])); |
| 794 | speex_bits_advance(bits, inband_skip_table[submode]); |
| 795 | } else if (submode == 13) { |
| 796 | /* user in-band; next 5 bits contain msg len */ |
| 797 | submode = speex_bits_unpack_unsigned(bits, 5); |
| 798 | TRACE__((THIS_FUNC, "Found submode: user-band %d bytes", submode)); |
| 799 | speex_bits_advance(bits, submode * 8); |
| 800 | } else if (submode > 8) { |
| 801 | TRACE__((THIS_FUNC, "Unknown sub-mode %d", submode)); |
| 802 | return -1; |
| 803 | } else { |
| 804 | /* NB frame */ |
| 805 | unsigned int advance = submode; |
| 806 | speex_mode_query(&speex_nb_mode, SPEEX_SUBMODE_BITS_PER_FRAME, &advance); |
| 807 | if (advance < 0) { |
| 808 | TRACE__((THIS_FUNC, "Invalid mode encountered. " |
| 809 | "The stream is corrupted.")); |
| 810 | return -1; |
| 811 | } |
| 812 | TRACE__((THIS_FUNC, "Submode %d: %d bits", submode, advance)); |
| 813 | advance -= (NB_SUBMODE_BITS+1); |
| 814 | speex_bits_advance(bits, advance); |
| 815 | |
| 816 | ++nb_count; |
| 817 | } |
| 818 | } |
| 819 | |
| 820 | return 0; |
| 821 | } |
| 822 | |
| 823 | |
| 824 | /* |
| 825 | * Get frames in the packet. |
| 826 | */ |
| 827 | static pj_status_t spx_codec_parse( pjmedia_codec *codec, |
| 828 | void *pkt, |
| 829 | pj_size_t pkt_size, |
| 830 | const pj_timestamp *ts, |
| 831 | unsigned *frame_cnt, |
| 832 | pjmedia_frame frames[]) |
| 833 | { |
| 834 | struct spx_private *spx = (struct spx_private*) codec->codec_data; |
| 835 | unsigned samples_per_frame; |
| 836 | unsigned count = 0; |
| 837 | int char_ptr = 0; |
| 838 | int bit_ptr = 0; |
| 839 | |
| 840 | samples_per_frame=spx_factory.speex_param[spx->param_id].samples_per_frame; |
| 841 | |
| 842 | /* Copy the data into the speex bit-stream */ |
| 843 | speex_bits_read_from(&spx->dec_bits, (char*)pkt, (int)pkt_size); |
| 844 | |
| 845 | while (speex_get_next_frame(&spx->dec_bits) == 0 && |
| 846 | spx->dec_bits.charPtr != char_ptr) |
| 847 | { |
| 848 | frames[count].buf = (char*)pkt + char_ptr; |
| 849 | /* Bit info contains start bit offset of the frame */ |
| 850 | frames[count].bit_info = bit_ptr; |
| 851 | frames[count].type = PJMEDIA_FRAME_TYPE_AUDIO; |
| 852 | frames[count].timestamp.u64 = ts->u64 + count * samples_per_frame; |
| 853 | frames[count].size = spx->dec_bits.charPtr - char_ptr; |
| 854 | if (spx->dec_bits.bitPtr) |
| 855 | ++frames[count].size; |
| 856 | |
| 857 | bit_ptr = spx->dec_bits.bitPtr; |
| 858 | char_ptr = spx->dec_bits.charPtr; |
| 859 | |
| 860 | ++count; |
| 861 | } |
| 862 | |
| 863 | *frame_cnt = count; |
| 864 | |
| 865 | return PJ_SUCCESS; |
| 866 | } |
| 867 | |
| 868 | /* |
| 869 | * Encode frames. |
| 870 | */ |
| 871 | static pj_status_t spx_codec_encode( pjmedia_codec *codec, |
| 872 | const struct pjmedia_frame *input, |
| 873 | unsigned output_buf_len, |
| 874 | struct pjmedia_frame *output) |
| 875 | { |
| 876 | struct spx_private *spx; |
| 877 | unsigned samples_per_frame; |
| 878 | int tx = 0; |
| 879 | spx_int16_t *pcm_in = (spx_int16_t*)input->buf; |
| 880 | pj_size_t nsamples; |
| 881 | |
| 882 | spx = (struct spx_private*) codec->codec_data; |
| 883 | |
| 884 | if (input->type != PJMEDIA_FRAME_TYPE_AUDIO) { |
| 885 | output->size = 0; |
| 886 | output->buf = NULL; |
| 887 | output->timestamp = input->timestamp; |
| 888 | output->type = input->type; |
| 889 | return PJ_SUCCESS; |
| 890 | } |
| 891 | |
| 892 | nsamples = input->size >> 1; |
| 893 | samples_per_frame=spx_factory.speex_param[spx->param_id].samples_per_frame; |
| 894 | |
| 895 | PJ_ASSERT_RETURN(nsamples % samples_per_frame == 0, |
| 896 | PJMEDIA_CODEC_EPCMFRMINLEN); |
| 897 | |
| 898 | /* Flush all the bits in the struct so we can encode a new frame */ |
| 899 | speex_bits_reset(&spx->enc_bits); |
| 900 | |
| 901 | /* Encode the frames */ |
| 902 | while (nsamples >= samples_per_frame) { |
| 903 | tx += speex_encode_int(spx->enc, pcm_in, &spx->enc_bits); |
| 904 | pcm_in += samples_per_frame; |
| 905 | nsamples -= samples_per_frame; |
| 906 | } |
| 907 | |
| 908 | /* Check if we need not to transmit the frame (DTX) */ |
| 909 | if (tx == 0) { |
| 910 | output->buf = NULL; |
| 911 | output->size = 0; |
| 912 | output->timestamp.u64 = input->timestamp.u64; |
| 913 | output->type = PJMEDIA_FRAME_TYPE_NONE; |
| 914 | return PJ_SUCCESS; |
| 915 | } |
| 916 | |
| 917 | /* Check size. */ |
| 918 | pj_assert(speex_bits_nbytes(&spx->enc_bits) <= (int)output_buf_len); |
| 919 | |
| 920 | /* Copy the bits to an array of char that can be written */ |
| 921 | output->size = speex_bits_write(&spx->enc_bits, |
| 922 | (char*)output->buf, output_buf_len); |
| 923 | output->type = PJMEDIA_FRAME_TYPE_AUDIO; |
| 924 | output->timestamp = input->timestamp; |
| 925 | |
| 926 | return PJ_SUCCESS; |
| 927 | } |
| 928 | |
| 929 | /* |
| 930 | * Decode frame. |
| 931 | */ |
| 932 | static pj_status_t spx_codec_decode( pjmedia_codec *codec, |
| 933 | const struct pjmedia_frame *input, |
| 934 | unsigned output_buf_len, |
| 935 | struct pjmedia_frame *output) |
| 936 | { |
| 937 | struct spx_private *spx; |
| 938 | unsigned samples_per_frame; |
| 939 | |
| 940 | spx = (struct spx_private*) codec->codec_data; |
| 941 | samples_per_frame=spx_factory.speex_param[spx->param_id].samples_per_frame; |
| 942 | |
| 943 | PJ_ASSERT_RETURN(output_buf_len >= samples_per_frame << 1, |
| 944 | PJMEDIA_CODEC_EPCMTOOSHORT); |
| 945 | |
| 946 | if (input->type != PJMEDIA_FRAME_TYPE_AUDIO) { |
| 947 | pjmedia_zero_samples((pj_int16_t*)output->buf, samples_per_frame); |
| 948 | output->size = samples_per_frame << 1; |
| 949 | output->timestamp.u64 = input->timestamp.u64; |
| 950 | output->type = PJMEDIA_FRAME_TYPE_AUDIO; |
| 951 | return PJ_SUCCESS; |
| 952 | } |
| 953 | |
| 954 | /* Copy the data into the bit-stream struct */ |
| 955 | speex_bits_read_from(&spx->dec_bits, (char*)input->buf, (int)input->size); |
| 956 | |
| 957 | /* Set Speex dec_bits pointer to the start bit of the frame */ |
| 958 | speex_bits_advance(&spx->dec_bits, input->bit_info); |
| 959 | |
| 960 | /* Decode the data */ |
| 961 | speex_decode_int(spx->dec, &spx->dec_bits, (spx_int16_t*)output->buf); |
| 962 | |
| 963 | output->type = PJMEDIA_FRAME_TYPE_AUDIO; |
| 964 | output->size = samples_per_frame << 1; |
| 965 | output->timestamp.u64 = input->timestamp.u64; |
| 966 | |
| 967 | return PJ_SUCCESS; |
| 968 | } |
| 969 | |
| 970 | /* |
| 971 | * Recover lost frame. |
| 972 | */ |
| 973 | static pj_status_t spx_codec_recover(pjmedia_codec *codec, |
| 974 | unsigned output_buf_len, |
| 975 | struct pjmedia_frame *output) |
| 976 | { |
| 977 | struct spx_private *spx; |
| 978 | unsigned count; |
| 979 | |
| 980 | /* output_buf_len is unreferenced when building in Release mode */ |
| 981 | PJ_UNUSED_ARG(output_buf_len); |
| 982 | |
| 983 | spx = (struct spx_private*) codec->codec_data; |
| 984 | |
| 985 | count = spx_factory.speex_param[spx->param_id].clock_rate * 20 / 1000; |
| 986 | pj_assert(count <= output_buf_len/2); |
| 987 | |
| 988 | /* Recover packet loss */ |
| 989 | speex_decode_int(spx->dec, NULL, (spx_int16_t*) output->buf); |
| 990 | |
| 991 | output->size = count * 2; |
| 992 | |
| 993 | return PJ_SUCCESS; |
| 994 | } |
| 995 | |
| 996 | |
| 997 | #endif /* PJMEDIA_HAS_SPEEX_CODEC */ |