1 /* GStreamer AAC parser plugin
2 * Copyright (C) 2008 Nokia Corporation. All rights reserved.
4 * Contact: Stefan Kost <stefan.kost@nokia.com>
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Library General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
11 * This library 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 GNU
14 * Library General Public License for more details.
16 * You should have received a copy of the GNU Library General Public
17 * License along with this library; if not, write to the
18 * Free Software Foundation, Inc., 51 Franklin St, Fifth Floor,
19 * Boston, MA 02110-1301, USA.
23 * SECTION:element-aacparse
24 * @short_description: AAC parser
25 * @see_also: #GstAmrParse
27 * This is an AAC parser which handles both ADIF and ADTS stream formats.
29 * As ADIF format is not framed, it is not seekable and stream duration cannot
30 * be determined either. However, ADTS format AAC clips can be seeked, and parser
31 * can also estimate playback position and clip duration.
34 * <title>Example launch line</title>
36 * gst-launch-1.0 filesrc location=abc.aac ! aacparse ! faad ! audioresample ! audioconvert ! alsasink
47 #include <gst/base/gstbitreader.h>
48 #include <gst/pbutils/pbutils.h>
49 #include "gstaacparse.h"
52 static GstStaticPadTemplate src_template = GST_STATIC_PAD_TEMPLATE ("src",
55 GST_STATIC_CAPS ("audio/mpeg, "
56 "framed = (boolean) true, " "mpegversion = (int) { 2, 4 }, "
57 "stream-format = (string) { raw, adts, adif, loas };"));
59 static GstStaticPadTemplate sink_template = GST_STATIC_PAD_TEMPLATE ("sink",
62 GST_STATIC_CAPS ("audio/mpeg, mpegversion = (int) { 2, 4 };"));
64 GST_DEBUG_CATEGORY_STATIC (aacparse_debug);
65 #define GST_CAT_DEFAULT aacparse_debug
68 #define ADIF_MAX_SIZE 40 /* Should be enough */
69 #define ADTS_MAX_SIZE 10 /* Should be enough */
70 #define LOAS_MAX_SIZE 3 /* Should be enough */
72 #define ADTS_HEADERS_LENGTH 7UL /* Total byte-length of fixed and variable
73 headers prepended during raw to ADTS
76 #define AAC_FRAME_DURATION(parse) (GST_SECOND/parse->frames_per_sec)
78 static const gint loas_sample_rate_table[32] = {
79 96000, 88200, 64000, 48000, 44100, 32000, 24000, 22050,
80 16000, 12000, 11025, 8000, 7350, 0, 0, 0
83 static const gint loas_channels_table[32] = {
84 0, 1, 2, 3, 4, 5, 6, 8,
85 0, 0, 0, 0, 0, 0, 0, 0
88 static gboolean gst_aac_parse_start (GstBaseParse * parse);
89 static gboolean gst_aac_parse_stop (GstBaseParse * parse);
91 static gboolean gst_aac_parse_sink_setcaps (GstBaseParse * parse,
93 static GstCaps *gst_aac_parse_sink_getcaps (GstBaseParse * parse,
96 static GstFlowReturn gst_aac_parse_handle_frame (GstBaseParse * parse,
97 GstBaseParseFrame * frame, gint * skipsize);
98 static GstFlowReturn gst_aac_parse_pre_push_frame (GstBaseParse * parse,
99 GstBaseParseFrame * frame);
101 G_DEFINE_TYPE (GstAacParse, gst_aac_parse, GST_TYPE_BASE_PARSE);
104 * gst_aac_parse_class_init:
105 * @klass: #GstAacParseClass.
109 gst_aac_parse_class_init (GstAacParseClass * klass)
111 GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
112 GstBaseParseClass *parse_class = GST_BASE_PARSE_CLASS (klass);
114 GST_DEBUG_CATEGORY_INIT (aacparse_debug, "aacparse", 0,
115 "AAC audio stream parser");
117 gst_element_class_add_pad_template (element_class,
118 gst_static_pad_template_get (&sink_template));
119 gst_element_class_add_pad_template (element_class,
120 gst_static_pad_template_get (&src_template));
122 gst_element_class_set_static_metadata (element_class,
123 "AAC audio stream parser", "Codec/Parser/Audio",
124 "Advanced Audio Coding parser", "Stefan Kost <stefan.kost@nokia.com>");
126 parse_class->start = GST_DEBUG_FUNCPTR (gst_aac_parse_start);
127 parse_class->stop = GST_DEBUG_FUNCPTR (gst_aac_parse_stop);
128 parse_class->set_sink_caps = GST_DEBUG_FUNCPTR (gst_aac_parse_sink_setcaps);
129 parse_class->get_sink_caps = GST_DEBUG_FUNCPTR (gst_aac_parse_sink_getcaps);
130 parse_class->handle_frame = GST_DEBUG_FUNCPTR (gst_aac_parse_handle_frame);
131 parse_class->pre_push_frame =
132 GST_DEBUG_FUNCPTR (gst_aac_parse_pre_push_frame);
137 * gst_aac_parse_init:
138 * @aacparse: #GstAacParse.
139 * @klass: #GstAacParseClass.
143 gst_aac_parse_init (GstAacParse * aacparse)
145 GST_DEBUG ("initialized");
146 GST_PAD_SET_ACCEPT_INTERSECT (GST_BASE_PARSE_SINK_PAD (aacparse));
151 * gst_aac_parse_set_src_caps:
152 * @aacparse: #GstAacParse.
153 * @sink_caps: (proposed) caps of sink pad
155 * Set source pad caps according to current knowledge about the
158 * Returns: TRUE if caps were successfully set.
161 gst_aac_parse_set_src_caps (GstAacParse * aacparse, GstCaps * sink_caps)
164 GstCaps *src_caps = NULL, *allowed;
165 gboolean res = FALSE;
166 const gchar *stream_format;
167 GstBuffer *codec_data;
168 guint16 codec_data_data;
170 GST_DEBUG_OBJECT (aacparse, "sink caps: %" GST_PTR_FORMAT, sink_caps);
172 src_caps = gst_caps_copy (sink_caps);
174 src_caps = gst_caps_new_empty_simple ("audio/mpeg");
176 gst_caps_set_simple (src_caps, "framed", G_TYPE_BOOLEAN, TRUE,
177 "mpegversion", G_TYPE_INT, aacparse->mpegversion, NULL);
179 aacparse->output_header_type = aacparse->header_type;
180 switch (aacparse->header_type) {
181 case DSPAAC_HEADER_NONE:
182 stream_format = "raw";
184 case DSPAAC_HEADER_ADTS:
185 stream_format = "adts";
187 case DSPAAC_HEADER_ADIF:
188 stream_format = "adif";
190 case DSPAAC_HEADER_LOAS:
191 stream_format = "loas";
194 stream_format = NULL;
197 s = gst_caps_get_structure (src_caps, 0);
198 if (aacparse->sample_rate > 0)
199 gst_structure_set (s, "rate", G_TYPE_INT, aacparse->sample_rate, NULL);
200 if (aacparse->channels > 0)
201 gst_structure_set (s, "channels", G_TYPE_INT, aacparse->channels, NULL);
203 gst_structure_set (s, "stream-format", G_TYPE_STRING, stream_format, NULL);
205 allowed = gst_pad_get_allowed_caps (GST_BASE_PARSE (aacparse)->srcpad);
206 if (!gst_caps_can_intersect (src_caps, allowed)) {
207 GST_DEBUG_OBJECT (GST_BASE_PARSE (aacparse)->srcpad,
208 "Caps can not intersect");
209 if (aacparse->header_type == DSPAAC_HEADER_ADTS) {
210 GST_DEBUG_OBJECT (GST_BASE_PARSE (aacparse)->srcpad,
211 "Input is ADTS, trying raw");
212 gst_caps_set_simple (src_caps, "stream-format", G_TYPE_STRING, "raw",
214 if (gst_caps_can_intersect (src_caps, allowed)) {
219 gst_codec_utils_aac_get_index_from_sample_rate
220 (aacparse->sample_rate);
222 goto not_a_known_rate;
224 GST_DEBUG_OBJECT (GST_BASE_PARSE (aacparse)->srcpad,
225 "Caps can intersect, we will drop the ADTS layer");
226 aacparse->output_header_type = DSPAAC_HEADER_NONE;
228 /* The codec_data data is according to AudioSpecificConfig,
229 ISO/IEC 14496-3, 1.6.2.1 */
230 codec_data = gst_buffer_new_and_alloc (2);
231 gst_buffer_map (codec_data, &map, GST_MAP_WRITE);
233 (aacparse->object_type << 11) |
234 (idx << 7) | (aacparse->channels << 3);
235 GST_WRITE_UINT16_BE (map.data, codec_data_data);
236 gst_buffer_unmap (codec_data, &map);
237 gst_caps_set_simple (src_caps, "codec_data", GST_TYPE_BUFFER,
240 } else if (aacparse->header_type == DSPAAC_HEADER_NONE) {
241 GST_DEBUG_OBJECT (GST_BASE_PARSE (aacparse)->srcpad,
242 "Input is raw, trying ADTS");
243 gst_caps_set_simple (src_caps, "stream-format", G_TYPE_STRING, "adts",
245 if (gst_caps_can_intersect (src_caps, allowed)) {
246 GST_DEBUG_OBJECT (GST_BASE_PARSE (aacparse)->srcpad,
247 "Caps can intersect, we will prepend ADTS headers");
248 aacparse->output_header_type = DSPAAC_HEADER_ADTS;
252 gst_caps_unref (allowed);
254 GST_DEBUG_OBJECT (aacparse, "setting src caps: %" GST_PTR_FORMAT, src_caps);
256 res = gst_pad_set_caps (GST_BASE_PARSE (aacparse)->srcpad, src_caps);
257 gst_caps_unref (src_caps);
261 gst_caps_unref (allowed);
262 gst_caps_unref (src_caps);
268 * gst_aac_parse_sink_setcaps:
272 * Implementation of "set_sink_caps" vmethod in #GstBaseParse class.
274 * Returns: TRUE on success.
277 gst_aac_parse_sink_setcaps (GstBaseParse * parse, GstCaps * caps)
279 GstAacParse *aacparse;
280 GstStructure *structure;
284 aacparse = GST_AAC_PARSE (parse);
285 structure = gst_caps_get_structure (caps, 0);
286 caps_str = gst_caps_to_string (caps);
288 GST_DEBUG_OBJECT (aacparse, "setcaps: %s", caps_str);
291 /* This is needed at least in case of RTP
292 * Parses the codec_data information to get ObjectType,
293 * number of channels and samplerate */
294 value = gst_structure_get_value (structure, "codec_data");
296 GstBuffer *buf = gst_value_get_buffer (value);
302 gst_buffer_map (buf, &map, GST_MAP_READ);
304 sr_idx = ((map.data[0] & 0x07) << 1) | ((map.data[1] & 0x80) >> 7);
305 aacparse->object_type = (map.data[0] & 0xf8) >> 3;
306 aacparse->sample_rate =
307 gst_codec_utils_aac_get_sample_rate_from_index (sr_idx);
308 aacparse->channels = (map.data[1] & 0x78) >> 3;
309 aacparse->header_type = DSPAAC_HEADER_NONE;
310 aacparse->mpegversion = 4;
311 aacparse->frame_samples = (map.data[1] & 4) ? 960 : 1024;
312 gst_buffer_unmap (buf, &map);
314 GST_DEBUG ("codec_data: object_type=%d, sample_rate=%d, channels=%d, "
315 "samples=%d", aacparse->object_type, aacparse->sample_rate,
316 aacparse->channels, aacparse->frame_samples);
318 /* arrange for metadata and get out of the way */
319 gst_aac_parse_set_src_caps (aacparse, caps);
320 if (aacparse->header_type == aacparse->output_header_type)
321 gst_base_parse_set_passthrough (parse, TRUE);
325 /* caps info overrides */
326 gst_structure_get_int (structure, "rate", &aacparse->sample_rate);
327 gst_structure_get_int (structure, "channels", &aacparse->channels);
329 aacparse->sample_rate = 0;
330 aacparse->channels = 0;
331 aacparse->header_type = DSPAAC_HEADER_NOT_PARSED;
332 gst_base_parse_set_passthrough (parse, FALSE);
340 * gst_aac_parse_adts_get_frame_len:
341 * @data: block of data containing an ADTS header.
343 * This function calculates ADTS frame length from the given header.
345 * Returns: size of the ADTS frame.
348 gst_aac_parse_adts_get_frame_len (const guint8 * data)
350 return ((data[3] & 0x03) << 11) | (data[4] << 3) | ((data[5] & 0xe0) >> 5);
355 * gst_aac_parse_check_adts_frame:
356 * @aacparse: #GstAacParse.
357 * @data: Data to be checked.
358 * @avail: Amount of data passed.
359 * @framesize: If valid ADTS frame was found, this will be set to tell the
360 * found frame size in bytes.
361 * @needed_data: If frame was not found, this may be set to tell how much
362 * more data is needed in the next round to detect the frame
363 * reliably. This may happen when a frame header candidate
364 * is found but it cannot be guaranteed to be the header without
365 * peeking the following data.
367 * Check if the given data contains contains ADTS frame. The algorithm
368 * will examine ADTS frame header and calculate the frame size. Also, another
369 * consecutive ADTS frame header need to be present after the found frame.
370 * Otherwise the data is not considered as a valid ADTS frame. However, this
371 * "extra check" is omitted when EOS has been received. In this case it is
372 * enough when data[0] contains a valid ADTS header.
374 * This function may set the #needed_data to indicate that a possible frame
375 * candidate has been found, but more data (#needed_data bytes) is needed to
376 * be absolutely sure. When this situation occurs, FALSE will be returned.
378 * When a valid frame is detected, this function will use
379 * gst_base_parse_set_min_frame_size() function from #GstBaseParse class
380 * to set the needed bytes for next frame.This way next data chunk is already
383 * Returns: TRUE if the given data contains a valid ADTS header.
386 gst_aac_parse_check_adts_frame (GstAacParse * aacparse,
387 const guint8 * data, const guint avail, gboolean drain,
388 guint * framesize, guint * needed_data)
392 if (G_UNLIKELY (avail < 2))
395 if ((data[0] == 0xff) && ((data[1] & 0xf6) == 0xf0)) {
397 /* This looks like an ADTS frame header but
398 we need at least 6 bytes to proceed */
399 if (G_UNLIKELY (avail < 6)) {
404 *framesize = gst_aac_parse_adts_get_frame_len (data);
406 /* In EOS mode this is enough. No need to examine the data further.
407 We also relax the check when we have sync, on the assumption that
408 if we're not looking at random data, we have a much higher chance
409 to get the correct sync, and this avoids losing two frames when
410 a single bit corruption happens. */
411 if (drain || !GST_BASE_PARSE_LOST_SYNC (aacparse)) {
415 if (*framesize + ADTS_MAX_SIZE > avail) {
416 /* We have found a possible frame header candidate, but can't be
417 sure since we don't have enough data to check the next frame */
418 GST_DEBUG ("NEED MORE DATA: we need %d, available %d",
419 *framesize + ADTS_MAX_SIZE, avail);
420 *needed_data = *framesize + ADTS_MAX_SIZE;
421 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (aacparse),
422 *framesize + ADTS_MAX_SIZE);
426 if ((data[*framesize] == 0xff) && ((data[*framesize + 1] & 0xf6) == 0xf0)) {
427 guint nextlen = gst_aac_parse_adts_get_frame_len (data + (*framesize));
429 GST_LOG ("ADTS frame found, len: %d bytes", *framesize);
430 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (aacparse),
431 nextlen + ADTS_MAX_SIZE);
439 gst_aac_parse_latm_get_value (GstAacParse * aacparse, GstBitReader * br,
442 guint8 bytes, i, byte;
445 if (!gst_bit_reader_get_bits_uint8 (br, &bytes, 2))
447 for (i = 0; i < bytes; ++i) {
449 if (!gst_bit_reader_get_bits_uint8 (br, &byte, 8))
457 gst_aac_parse_get_audio_object_type (GstAacParse * aacparse, GstBitReader * br,
458 guint8 * audio_object_type)
460 if (!gst_bit_reader_get_bits_uint8 (br, audio_object_type, 5))
462 if (*audio_object_type == 31) {
463 if (!gst_bit_reader_get_bits_uint8 (br, audio_object_type, 6))
465 *audio_object_type += 32;
467 GST_LOG_OBJECT (aacparse, "audio object type %u", *audio_object_type);
472 gst_aac_parse_get_audio_sample_rate (GstAacParse * aacparse, GstBitReader * br,
475 guint8 sampling_frequency_index;
476 if (!gst_bit_reader_get_bits_uint8 (br, &sampling_frequency_index, 4))
478 GST_LOG_OBJECT (aacparse, "sampling_frequency_index: %u",
479 sampling_frequency_index);
480 if (sampling_frequency_index == 0xf) {
481 guint32 sampling_rate;
482 if (!gst_bit_reader_get_bits_uint32 (br, &sampling_rate, 24))
484 *sample_rate = sampling_rate;
486 *sample_rate = loas_sample_rate_table[sampling_frequency_index];
493 /* See table 1.13 in ISO/IEC 14496-3 */
495 gst_aac_parse_read_loas_audio_specific_config (GstAacParse * aacparse,
496 GstBitReader * br, gint * sample_rate, gint * channels, guint32 * bits)
498 guint8 audio_object_type, channel_configuration;
500 if (!gst_aac_parse_get_audio_object_type (aacparse, br, &audio_object_type))
503 if (!gst_aac_parse_get_audio_sample_rate (aacparse, br, sample_rate))
506 if (!gst_bit_reader_get_bits_uint8 (br, &channel_configuration, 4))
508 GST_LOG_OBJECT (aacparse, "channel_configuration: %d", channel_configuration);
509 *channels = loas_channels_table[channel_configuration];
513 if (audio_object_type == 5) {
514 GST_LOG_OBJECT (aacparse,
515 "Audio object type 5, so rereading sampling rate...");
516 if (!gst_aac_parse_get_audio_sample_rate (aacparse, br, sample_rate))
520 GST_INFO_OBJECT (aacparse, "Found LOAS config: %d Hz, %d channels",
521 *sample_rate, *channels);
523 /* There's LOTS of stuff next, but we ignore it for now as we have
524 what we want (sample rate and number of channels */
525 GST_DEBUG_OBJECT (aacparse,
526 "Need more code to parse humongous LOAS data, currently ignored");
534 gst_aac_parse_read_loas_config (GstAacParse * aacparse, const guint8 * data,
535 guint avail, gint * sample_rate, gint * channels, gint * version)
540 /* No version in the bitstream, but the spec has LOAS in the MPEG-4 section */
544 gst_bit_reader_init (&br, data, avail);
546 /* skip sync word (11 bits) and size (13 bits) */
547 if (!gst_bit_reader_skip (&br, 11 + 13))
550 /* First bit is "use last config" */
551 if (!gst_bit_reader_get_bits_uint8 (&br, &u8, 1))
554 GST_DEBUG_OBJECT (aacparse, "Frame uses previous config");
555 if (!aacparse->sample_rate || !aacparse->channels) {
556 GST_WARNING_OBJECT (aacparse, "No previous config to use");
558 *sample_rate = aacparse->sample_rate;
559 *channels = aacparse->channels;
563 GST_DEBUG_OBJECT (aacparse, "Frame contains new config");
565 if (!gst_bit_reader_get_bits_uint8 (&br, &v, 1))
568 if (!gst_bit_reader_get_bits_uint8 (&br, &vA, 1))
573 GST_LOG_OBJECT (aacparse, "v %d, vA %d", v, vA);
575 guint8 same_time, subframes, num_program, prog;
578 if (!gst_aac_parse_latm_get_value (aacparse, &br, &value))
581 if (!gst_bit_reader_get_bits_uint8 (&br, &same_time, 1))
583 if (!gst_bit_reader_get_bits_uint8 (&br, &subframes, 6))
585 if (!gst_bit_reader_get_bits_uint8 (&br, &num_program, 4))
587 GST_LOG_OBJECT (aacparse, "same_time %d, subframes %d, num_program %d",
588 same_time, subframes, num_program);
590 for (prog = 0; prog <= num_program; ++prog) {
591 guint8 num_layer, layer;
592 if (!gst_bit_reader_get_bits_uint8 (&br, &num_layer, 3))
594 GST_LOG_OBJECT (aacparse, "Program %d: %d layers", prog, num_layer);
596 for (layer = 0; layer <= num_layer; ++layer) {
597 guint8 use_same_config;
598 if (prog == 0 && layer == 0) {
601 if (!gst_bit_reader_get_bits_uint8 (&br, &use_same_config, 1))
604 if (!use_same_config) {
606 if (!gst_aac_parse_read_loas_audio_specific_config (aacparse, &br,
607 sample_rate, channels, NULL))
610 guint32 bits, asc_len;
611 if (!gst_aac_parse_latm_get_value (aacparse, &br, &asc_len))
613 if (!gst_aac_parse_read_loas_audio_specific_config (aacparse, &br,
614 sample_rate, channels, &bits))
617 if (!gst_bit_reader_skip (&br, asc_len))
623 GST_LOG_OBJECT (aacparse, "More data ignored");
625 GST_WARNING_OBJECT (aacparse, "Spec says \"TBD\"...");
631 * gst_aac_parse_loas_get_frame_len:
632 * @data: block of data containing a LOAS header.
634 * This function calculates LOAS frame length from the given header.
636 * Returns: size of the LOAS frame.
639 gst_aac_parse_loas_get_frame_len (const guint8 * data)
641 return (((data[1] & 0x1f) << 8) | data[2]) + 3;
646 * gst_aac_parse_check_loas_frame:
647 * @aacparse: #GstAacParse.
648 * @data: Data to be checked.
649 * @avail: Amount of data passed.
650 * @framesize: If valid LOAS frame was found, this will be set to tell the
651 * found frame size in bytes.
652 * @needed_data: If frame was not found, this may be set to tell how much
653 * more data is needed in the next round to detect the frame
654 * reliably. This may happen when a frame header candidate
655 * is found but it cannot be guaranteed to be the header without
656 * peeking the following data.
658 * Check if the given data contains contains LOAS frame. The algorithm
659 * will examine LOAS frame header and calculate the frame size. Also, another
660 * consecutive LOAS frame header need to be present after the found frame.
661 * Otherwise the data is not considered as a valid LOAS frame. However, this
662 * "extra check" is omitted when EOS has been received. In this case it is
663 * enough when data[0] contains a valid LOAS header.
665 * This function may set the #needed_data to indicate that a possible frame
666 * candidate has been found, but more data (#needed_data bytes) is needed to
667 * be absolutely sure. When this situation occurs, FALSE will be returned.
669 * When a valid frame is detected, this function will use
670 * gst_base_parse_set_min_frame_size() function from #GstBaseParse class
671 * to set the needed bytes for next frame.This way next data chunk is already
674 * LOAS can have three different formats, if I read the spec correctly. Only
675 * one of them is supported here, as the two samples I have use this one.
677 * Returns: TRUE if the given data contains a valid LOAS header.
680 gst_aac_parse_check_loas_frame (GstAacParse * aacparse,
681 const guint8 * data, const guint avail, gboolean drain,
682 guint * framesize, guint * needed_data)
687 if (G_UNLIKELY (avail < 3))
690 if ((data[0] == 0x56) && ((data[1] & 0xe0) == 0xe0)) {
691 *framesize = gst_aac_parse_loas_get_frame_len (data);
692 GST_DEBUG_OBJECT (aacparse, "Found %u byte LOAS frame", *framesize);
694 /* In EOS mode this is enough. No need to examine the data further.
695 We also relax the check when we have sync, on the assumption that
696 if we're not looking at random data, we have a much higher chance
697 to get the correct sync, and this avoids losing two frames when
698 a single bit corruption happens. */
699 if (drain || !GST_BASE_PARSE_LOST_SYNC (aacparse)) {
703 if (*framesize + LOAS_MAX_SIZE > avail) {
704 /* We have found a possible frame header candidate, but can't be
705 sure since we don't have enough data to check the next frame */
706 GST_DEBUG ("NEED MORE DATA: we need %d, available %d",
707 *framesize + LOAS_MAX_SIZE, avail);
708 *needed_data = *framesize + LOAS_MAX_SIZE;
709 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (aacparse),
710 *framesize + LOAS_MAX_SIZE);
714 if ((data[*framesize] == 0x56) && ((data[*framesize + 1] & 0xe0) == 0xe0)) {
715 guint nextlen = gst_aac_parse_loas_get_frame_len (data + (*framesize));
717 GST_LOG ("LOAS frame found, len: %d bytes", *framesize);
718 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (aacparse),
719 nextlen + LOAS_MAX_SIZE);
726 /* caller ensure sufficient data */
728 gst_aac_parse_parse_adts_header (GstAacParse * aacparse, const guint8 * data,
729 gint * rate, gint * channels, gint * object, gint * version)
733 gint sr_idx = (data[2] & 0x3c) >> 2;
735 *rate = gst_codec_utils_aac_get_sample_rate_from_index (sr_idx);
738 *channels = ((data[2] & 0x01) << 2) | ((data[3] & 0xc0) >> 6);
741 *version = (data[1] & 0x08) ? 2 : 4;
743 *object = ((data[2] & 0xc0) >> 6) + 1;
747 * gst_aac_parse_detect_stream:
748 * @aacparse: #GstAacParse.
749 * @data: A block of data that needs to be examined for stream characteristics.
750 * @avail: Size of the given datablock.
751 * @framesize: If valid stream was found, this will be set to tell the
752 * first frame size in bytes.
753 * @skipsize: If valid stream was found, this will be set to tell the first
754 * audio frame position within the given data.
756 * Examines the given piece of data and try to detect the format of it. It
757 * checks for "ADIF" header (in the beginning of the clip) and ADTS frame
758 * header. If the stream is detected, TRUE will be returned and #framesize
759 * is set to indicate the found frame size. Additionally, #skipsize might
760 * be set to indicate the number of bytes that need to be skipped, a.k.a. the
761 * position of the frame inside given data chunk.
763 * Returns: TRUE on success.
766 gst_aac_parse_detect_stream (GstAacParse * aacparse,
767 const guint8 * data, const guint avail, gboolean drain,
768 guint * framesize, gint * skipsize)
770 gboolean found = FALSE;
771 guint need_data_adts = 0, need_data_loas;
774 GST_DEBUG_OBJECT (aacparse, "Parsing header data");
776 /* FIXME: No need to check for ADIF if we are not in the beginning of the
779 /* Can we even parse the header? */
780 if (avail < MAX (ADTS_MAX_SIZE, LOAS_MAX_SIZE)) {
781 GST_DEBUG_OBJECT (aacparse, "Not enough data to check");
785 for (i = 0; i < avail - 4; i++) {
786 if (((data[i] == 0xff) && ((data[i + 1] & 0xf6) == 0xf0)) ||
787 ((data[0] == 0x56) && ((data[1] & 0xe0) == 0xe0)) ||
788 strncmp ((char *) data + i, "ADIF", 4) == 0) {
789 GST_DEBUG_OBJECT (aacparse, "Found signature at offset %u", i);
793 /* Trick: tell the parent class that we didn't find the frame yet,
794 but make it skip 'i' amount of bytes. Next time we arrive
795 here we have full frame in the beginning of the data. */
808 if (gst_aac_parse_check_adts_frame (aacparse, data, avail, drain,
809 framesize, &need_data_adts)) {
812 GST_INFO ("ADTS ID: %d, framesize: %d", (data[1] & 0x08) >> 3, *framesize);
814 gst_aac_parse_parse_adts_header (aacparse, data, &rate, &channels,
815 &aacparse->object_type, &aacparse->mpegversion);
817 if (!channels || !framesize) {
818 GST_DEBUG_OBJECT (aacparse, "impossible ADTS configuration");
822 aacparse->header_type = DSPAAC_HEADER_ADTS;
823 gst_base_parse_set_frame_rate (GST_BASE_PARSE (aacparse), rate,
824 aacparse->frame_samples, 2, 2);
826 GST_DEBUG ("ADTS: samplerate %d, channels %d, objtype %d, version %d",
827 rate, channels, aacparse->object_type, aacparse->mpegversion);
829 gst_base_parse_set_syncable (GST_BASE_PARSE (aacparse), TRUE);
834 if (gst_aac_parse_check_loas_frame (aacparse, data, avail, drain,
835 framesize, &need_data_loas)) {
838 GST_INFO ("LOAS, framesize: %d", *framesize);
840 aacparse->header_type = DSPAAC_HEADER_LOAS;
842 if (!gst_aac_parse_read_loas_config (aacparse, data, avail, &rate,
843 &channels, &aacparse->mpegversion)) {
844 GST_WARNING_OBJECT (aacparse, "Error reading LOAS config");
848 if (rate && channels) {
849 gst_base_parse_set_frame_rate (GST_BASE_PARSE (aacparse), rate,
850 aacparse->frame_samples, 2, 2);
852 GST_DEBUG ("LOAS: samplerate %d, channels %d, objtype %d, version %d",
853 rate, channels, aacparse->object_type, aacparse->mpegversion);
854 aacparse->sample_rate = rate;
855 aacparse->channels = channels;
858 gst_base_parse_set_syncable (GST_BASE_PARSE (aacparse), TRUE);
863 if (need_data_adts || need_data_loas) {
864 /* This tells the parent class not to skip any data */
869 if (avail < ADIF_MAX_SIZE)
872 if (memcmp (data + i, "ADIF", 4) == 0) {
879 aacparse->header_type = DSPAAC_HEADER_ADIF;
880 aacparse->mpegversion = 4;
882 /* Skip the "ADIF" bytes */
885 /* copyright string */
887 skip_size += 9; /* skip 9 bytes */
889 bitstream_type = adif[0 + skip_size] & 0x10;
891 ((unsigned int) (adif[0 + skip_size] & 0x0f) << 19) |
892 ((unsigned int) adif[1 + skip_size] << 11) |
893 ((unsigned int) adif[2 + skip_size] << 3) |
894 ((unsigned int) adif[3 + skip_size] & 0xe0);
897 if (bitstream_type == 0) {
899 /* Buffer fullness parsing. Currently not needed... */
903 num_elems = (adif[3 + skip_size] & 0x1e);
904 GST_INFO ("ADIF num_config_elems: %d", num_elems);
906 fullness = ((unsigned int) (adif[3 + skip_size] & 0x01) << 19) |
907 ((unsigned int) adif[4 + skip_size] << 11) |
908 ((unsigned int) adif[5 + skip_size] << 3) |
909 ((unsigned int) (adif[6 + skip_size] & 0xe0) >> 5);
911 GST_INFO ("ADIF buffer fullness: %d", fullness);
913 aacparse->object_type = ((adif[6 + skip_size] & 0x01) << 1) |
914 ((adif[7 + skip_size] & 0x80) >> 7);
915 sr_idx = (adif[7 + skip_size] & 0x78) >> 3;
919 aacparse->object_type = (adif[4 + skip_size] & 0x18) >> 3;
920 sr_idx = ((adif[4 + skip_size] & 0x07) << 1) |
921 ((adif[5 + skip_size] & 0x80) >> 7);
924 /* FIXME: This gives totally wrong results. Duration calculation cannot
926 aacparse->sample_rate =
927 gst_codec_utils_aac_get_sample_rate_from_index (sr_idx);
929 /* baseparse is not given any fps,
930 * so it will give up on timestamps, seeking, etc */
932 /* FIXME: Can we assume this? */
933 aacparse->channels = 2;
935 GST_INFO ("ADIF: br=%d, samplerate=%d, objtype=%d",
936 aacparse->bitrate, aacparse->sample_rate, aacparse->object_type);
938 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (aacparse), 512);
940 /* arrange for metadata and get out of the way */
941 sinkcaps = gst_pad_get_current_caps (GST_BASE_PARSE_SINK_PAD (aacparse));
942 gst_aac_parse_set_src_caps (aacparse, sinkcaps);
944 gst_caps_unref (sinkcaps);
946 /* not syncable, not easily seekable (unless we push data from start */
947 gst_base_parse_set_syncable (GST_BASE_PARSE_CAST (aacparse), FALSE);
948 gst_base_parse_set_passthrough (GST_BASE_PARSE_CAST (aacparse), TRUE);
949 gst_base_parse_set_average_bitrate (GST_BASE_PARSE_CAST (aacparse), 0);
955 /* This should never happen */
960 * gst_aac_parse_get_audio_profile_object_type
961 * @aacparse: #GstAacParse.
963 * Gets the MPEG-2 profile or the MPEG-4 object type value corresponding to the
964 * mpegversion and profile of @aacparse's src pad caps, according to the
965 * values defined by table 1.A.11 in ISO/IEC 14496-3.
967 * Returns: the profile or object type value corresponding to @aacparse's src
968 * pad caps, if such a value exists; otherwise G_MAXUINT8.
971 gst_aac_parse_get_audio_profile_object_type (GstAacParse * aacparse)
974 GstStructure *srcstruct;
975 const gchar *profile;
978 srccaps = gst_pad_get_current_caps (GST_BASE_PARSE_SRC_PAD (aacparse));
979 srcstruct = gst_caps_get_structure (srccaps, 0);
980 profile = gst_structure_get_string (srcstruct, "profile");
981 if (G_UNLIKELY (profile == NULL)) {
982 gst_caps_unref (srccaps);
986 if (g_strcmp0 (profile, "main") == 0) {
988 } else if (g_strcmp0 (profile, "lc") == 0) {
990 } else if (g_strcmp0 (profile, "ssr") == 0) {
992 } else if (g_strcmp0 (profile, "ltp") == 0) {
993 if (G_LIKELY (aacparse->mpegversion == 4))
996 ret = G_MAXUINT8; /* LTP Object Type allowed only for MPEG-4 */
1001 gst_caps_unref (srccaps);
1006 * gst_aac_parse_get_audio_channel_configuration
1007 * @num_channels: number of audio channels.
1009 * Gets the Channel Configuration value, as defined by table 1.19 in ISO/IEC
1010 * 14496-3, for a given number of audio channels.
1012 * Returns: the Channel Configuration value corresponding to @num_channels, if
1013 * such a value exists; otherwise G_MAXUINT8.
1016 gst_aac_parse_get_audio_channel_configuration (gint num_channels)
1018 if (num_channels >= 1 && num_channels <= 6) /* Mono up to & including 5.1 */
1019 return (guint8) num_channels;
1020 else if (num_channels == 8) /* 7.1 */
1027 * gst_aac_parse_get_audio_sampling_frequency_index:
1028 * @sample_rate: audio sampling rate.
1030 * Gets the Sampling Frequency Index value, as defined by table 1.18 in ISO/IEC
1031 * 14496-3, for a given sampling rate.
1033 * Returns: the Sampling Frequency Index value corresponding to @sample_rate,
1034 * if such a value exists; otherwise G_MAXUINT8.
1037 gst_aac_parse_get_audio_sampling_frequency_index (gint sample_rate)
1039 switch (sample_rate) {
1072 * gst_aac_parse_prepend_adts_headers:
1073 * @aacparse: #GstAacParse.
1074 * @frame: raw AAC frame to which ADTS headers shall be prepended.
1076 * Prepends ADTS headers to a raw AAC audio frame.
1078 * Returns: TRUE if ADTS headers were successfully prepended; FALSE otherwise.
1081 gst_aac_parse_prepend_adts_headers (GstAacParse * aacparse,
1082 GstBaseParseFrame * frame)
1085 guint8 *adts_headers;
1088 guint8 id, profile, channel_configuration, sampling_frequency_index;
1090 id = (aacparse->mpegversion == 4) ? 0x0U : 0x1U;
1091 profile = gst_aac_parse_get_audio_profile_object_type (aacparse);
1092 if (profile == G_MAXUINT8) {
1093 GST_ERROR_OBJECT (aacparse, "Unsupported audio profile or object type");
1096 channel_configuration =
1097 gst_aac_parse_get_audio_channel_configuration (aacparse->channels);
1098 if (channel_configuration == G_MAXUINT8) {
1099 GST_ERROR_OBJECT (aacparse, "Unsupported number of channels");
1102 sampling_frequency_index =
1103 gst_aac_parse_get_audio_sampling_frequency_index (aacparse->sample_rate);
1104 if (sampling_frequency_index == G_MAXUINT8) {
1105 GST_ERROR_OBJECT (aacparse, "Unsupported sampling frequency");
1109 frame->out_buffer = gst_buffer_copy (frame->buffer);
1110 buf_size = gst_buffer_get_size (frame->out_buffer);
1111 frame_size = buf_size + ADTS_HEADERS_LENGTH;
1113 if (G_UNLIKELY (frame_size >= 0x4000)) {
1114 GST_ERROR_OBJECT (aacparse, "Frame size is too big for ADTS");
1118 adts_headers = (guint8 *) g_malloc0 (ADTS_HEADERS_LENGTH);
1120 /* Note: no error correction bits are added to the resulting ADTS frames */
1121 adts_headers[0] = 0xFFU;
1122 adts_headers[1] = 0xF0U | (id << 3) | 0x1U;
1123 adts_headers[2] = (profile << 6) | (sampling_frequency_index << 2) | 0x2U |
1124 (channel_configuration & 0x4U);
1125 adts_headers[3] = ((channel_configuration & 0x3U) << 6) | 0x30U |
1126 (guint8) (frame_size >> 11);
1127 adts_headers[4] = (guint8) ((frame_size >> 3) & 0x00FF);
1128 adts_headers[5] = (guint8) (((frame_size & 0x0007) << 5) + 0x1FU);
1129 adts_headers[6] = 0xFCU;
1131 mem = gst_memory_new_wrapped (0, adts_headers, ADTS_HEADERS_LENGTH, 0,
1132 ADTS_HEADERS_LENGTH, NULL, NULL);
1133 gst_buffer_prepend_memory (frame->out_buffer, mem);
1139 * gst_aac_parse_check_valid_frame:
1140 * @parse: #GstBaseParse.
1141 * @frame: #GstBaseParseFrame.
1142 * @skipsize: How much data parent class should skip in order to find the
1145 * Implementation of "handle_frame" vmethod in #GstBaseParse class.
1147 * Also determines frame overhead.
1148 * ADTS streams have a 7 byte header in each frame. MP4 and ADIF streams don't have
1149 * a per-frame header. LOAS has 3 bytes.
1151 * We're making a couple of simplifying assumptions:
1153 * 1. We count Program Configuration Elements rather than searching for them
1154 * in the streams to discount them - the overhead is negligible.
1156 * 2. We ignore CRC. This has a worst-case impact of (num_raw_blocks + 1)*16
1157 * bits, which should still not be significant enough to warrant the
1158 * additional parsing through the headers
1160 * Returns: a #GstFlowReturn.
1162 static GstFlowReturn
1163 gst_aac_parse_handle_frame (GstBaseParse * parse,
1164 GstBaseParseFrame * frame, gint * skipsize)
1167 GstAacParse *aacparse;
1168 gboolean ret = FALSE;
1172 gint rate, channels;
1174 aacparse = GST_AAC_PARSE (parse);
1175 buffer = frame->buffer;
1177 gst_buffer_map (buffer, &map, GST_MAP_READ);
1180 lost_sync = GST_BASE_PARSE_LOST_SYNC (parse);
1182 if (aacparse->header_type == DSPAAC_HEADER_ADIF ||
1183 aacparse->header_type == DSPAAC_HEADER_NONE) {
1184 /* There is nothing to parse */
1185 framesize = map.size;
1188 } else if (aacparse->header_type == DSPAAC_HEADER_NOT_PARSED || lost_sync) {
1190 ret = gst_aac_parse_detect_stream (aacparse, map.data, map.size,
1191 GST_BASE_PARSE_DRAINING (parse), &framesize, skipsize);
1193 } else if (aacparse->header_type == DSPAAC_HEADER_ADTS) {
1194 guint needed_data = 1024;
1196 ret = gst_aac_parse_check_adts_frame (aacparse, map.data, map.size,
1197 GST_BASE_PARSE_DRAINING (parse), &framesize, &needed_data);
1200 GST_DEBUG ("buffer didn't contain valid frame");
1201 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (aacparse),
1205 } else if (aacparse->header_type == DSPAAC_HEADER_LOAS) {
1206 guint needed_data = 1024;
1208 ret = gst_aac_parse_check_loas_frame (aacparse, map.data,
1209 map.size, GST_BASE_PARSE_DRAINING (parse), &framesize, &needed_data);
1212 GST_DEBUG ("buffer didn't contain valid frame");
1213 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (aacparse),
1218 GST_DEBUG ("buffer didn't contain valid frame");
1219 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (aacparse),
1223 if (G_UNLIKELY (!ret))
1226 if (aacparse->header_type == DSPAAC_HEADER_ADTS) {
1228 frame->overhead = 7;
1230 gst_aac_parse_parse_adts_header (aacparse, map.data,
1231 &rate, &channels, NULL, NULL);
1233 GST_LOG_OBJECT (aacparse, "rate: %d, chans: %d", rate, channels);
1235 if (G_UNLIKELY (rate != aacparse->sample_rate
1236 || channels != aacparse->channels)) {
1237 aacparse->sample_rate = rate;
1238 aacparse->channels = channels;
1240 if (!gst_aac_parse_set_src_caps (aacparse, NULL)) {
1241 /* If linking fails, we need to return appropriate error */
1242 ret = GST_FLOW_NOT_LINKED;
1245 gst_base_parse_set_frame_rate (GST_BASE_PARSE (aacparse),
1246 aacparse->sample_rate, aacparse->frame_samples, 2, 2);
1248 } else if (aacparse->header_type == DSPAAC_HEADER_LOAS) {
1249 gboolean setcaps = FALSE;
1252 frame->overhead = 3;
1254 if (!gst_aac_parse_read_loas_config (aacparse, map.data, map.size, &rate,
1256 GST_WARNING_OBJECT (aacparse, "Error reading LOAS config");
1257 } else if (G_UNLIKELY (rate != aacparse->sample_rate
1258 || channels != aacparse->channels)) {
1259 aacparse->sample_rate = rate;
1260 aacparse->channels = channels;
1262 GST_INFO_OBJECT (aacparse, "New LOAS config: %d Hz, %d channels", rate,
1266 /* We want to set caps both at start, and when rate/channels change.
1267 Since only some LOAS frames have that info, we may receive frames
1268 before knowing about rate/channels. */
1270 || !gst_pad_has_current_caps (GST_BASE_PARSE_SRC_PAD (aacparse))) {
1271 if (!gst_aac_parse_set_src_caps (aacparse, NULL)) {
1272 /* If linking fails, we need to return appropriate error */
1273 ret = GST_FLOW_NOT_LINKED;
1276 gst_base_parse_set_frame_rate (GST_BASE_PARSE (aacparse),
1277 aacparse->sample_rate, aacparse->frame_samples, 2, 2);
1281 if (aacparse->header_type == DSPAAC_HEADER_NONE
1282 && aacparse->output_header_type == DSPAAC_HEADER_ADTS) {
1283 if (!gst_aac_parse_prepend_adts_headers (aacparse, frame)) {
1284 GST_ERROR_OBJECT (aacparse, "Failed to prepend ADTS headers to frame");
1285 ret = GST_FLOW_ERROR;
1290 gst_buffer_unmap (buffer, &map);
1293 /* found, skip if needed */
1302 if (ret && framesize <= map.size) {
1303 return gst_base_parse_finish_frame (parse, frame, framesize);
1309 static GstFlowReturn
1310 gst_aac_parse_pre_push_frame (GstBaseParse * parse, GstBaseParseFrame * frame)
1312 GstAacParse *aacparse = GST_AAC_PARSE (parse);
1314 if (!aacparse->sent_codec_tag) {
1315 GstTagList *taglist;
1318 taglist = gst_tag_list_new_empty ();
1321 caps = gst_pad_get_current_caps (GST_BASE_PARSE_SRC_PAD (parse));
1322 gst_pb_utils_add_codec_description_to_tag_list (taglist,
1323 GST_TAG_AUDIO_CODEC, caps);
1324 gst_caps_unref (caps);
1326 gst_pad_push_event (GST_BASE_PARSE_SRC_PAD (aacparse),
1327 gst_event_new_tag (taglist));
1329 /* also signals the end of first-frame processing */
1330 aacparse->sent_codec_tag = TRUE;
1333 /* As a special case, we can remove the ADTS framing and output raw AAC. */
1334 if (aacparse->header_type == DSPAAC_HEADER_ADTS
1335 && aacparse->output_header_type == DSPAAC_HEADER_NONE) {
1338 gst_buffer_map (frame->buffer, &map, GST_MAP_READ);
1339 header_size = (map.data[1] & 1) ? 7 : 9; /* optional CRC */
1340 gst_buffer_unmap (frame->buffer, &map);
1341 gst_buffer_resize (frame->buffer, header_size,
1342 gst_buffer_get_size (frame->buffer) - header_size);
1350 * gst_aac_parse_start:
1351 * @parse: #GstBaseParse.
1353 * Implementation of "start" vmethod in #GstBaseParse class.
1355 * Returns: TRUE if startup succeeded.
1358 gst_aac_parse_start (GstBaseParse * parse)
1360 GstAacParse *aacparse;
1362 aacparse = GST_AAC_PARSE (parse);
1363 GST_DEBUG ("start");
1364 aacparse->frame_samples = 1024;
1365 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (aacparse), ADTS_MAX_SIZE);
1366 aacparse->sent_codec_tag = FALSE;
1372 * gst_aac_parse_stop:
1373 * @parse: #GstBaseParse.
1375 * Implementation of "stop" vmethod in #GstBaseParse class.
1377 * Returns: TRUE is stopping succeeded.
1380 gst_aac_parse_stop (GstBaseParse * parse)
1387 remove_fields (GstCaps * caps)
1391 n = gst_caps_get_size (caps);
1392 for (i = 0; i < n; i++) {
1393 GstStructure *s = gst_caps_get_structure (caps, i);
1395 gst_structure_remove_field (s, "framed");
1400 gst_aac_parse_sink_getcaps (GstBaseParse * parse, GstCaps * filter)
1402 GstCaps *peercaps, *templ;
1405 templ = gst_pad_get_pad_template_caps (GST_BASE_PARSE_SINK_PAD (parse));
1408 GstCaps *fcopy = gst_caps_copy (filter);
1409 /* Remove the fields we convert */
1410 remove_fields (fcopy);
1411 peercaps = gst_pad_peer_query_caps (GST_BASE_PARSE_SRC_PAD (parse), fcopy);
1412 gst_caps_unref (fcopy);
1414 peercaps = gst_pad_peer_query_caps (GST_BASE_PARSE_SRC_PAD (parse), NULL);
1417 peercaps = gst_caps_make_writable (peercaps);
1418 /* Remove the fields we convert */
1419 remove_fields (peercaps);
1421 res = gst_caps_intersect_full (peercaps, templ, GST_CAPS_INTERSECT_FIRST);
1422 gst_caps_unref (peercaps);
1423 gst_caps_unref (templ);
1429 GstCaps *intersection;
1432 gst_caps_intersect_full (filter, res, GST_CAPS_INTERSECT_FIRST);
1433 gst_caps_unref (res);