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., 59 Temple Place - Suite 330,
19 * Boston, MA 02111-1307, 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 filesrc location=abc.aac ! aacparse ! faad ! audioresample ! audioconvert ! alsasink
47 #include "gstaacparse.h"
50 static GstStaticPadTemplate src_template = GST_STATIC_PAD_TEMPLATE ("src",
53 GST_STATIC_CAPS ("audio/mpeg, "
54 "framed = (boolean) true, " "mpegversion = (int) { 2, 4 }, "
55 "stream-format = (string) { raw, adts, adif };"));
57 static GstStaticPadTemplate sink_template = GST_STATIC_PAD_TEMPLATE ("sink",
60 GST_STATIC_CAPS ("audio/mpeg, mpegversion = (int) { 2, 4 };"));
62 GST_DEBUG_CATEGORY_STATIC (aacparse_debug);
63 #define GST_CAT_DEFAULT aacparse_debug
66 #define ADIF_MAX_SIZE 40 /* Should be enough */
67 #define ADTS_MAX_SIZE 10 /* Should be enough */
70 #define AAC_FRAME_DURATION(parse) (GST_SECOND/parse->frames_per_sec)
72 static gboolean gst_aac_parse_start (GstBaseParse * parse);
73 static gboolean gst_aac_parse_stop (GstBaseParse * parse);
75 static gboolean gst_aac_parse_sink_setcaps (GstBaseParse * parse,
77 static GstCaps *gst_aac_parse_sink_getcaps (GstBaseParse * parse);
79 static gboolean gst_aac_parse_check_valid_frame (GstBaseParse * parse,
80 GstBaseParseFrame * frame, guint * size, gint * skipsize);
82 static GstFlowReturn gst_aac_parse_parse_frame (GstBaseParse * parse,
83 GstBaseParseFrame * frame);
85 #define _do_init(bla) \
86 GST_DEBUG_CATEGORY_INIT (aacparse_debug, "aacparse", 0, \
87 "AAC audio stream parser");
89 GST_BOILERPLATE_FULL (GstAacParse, gst_aac_parse, GstBaseParse,
90 GST_TYPE_BASE_PARSE, _do_init);
93 gst_aac_parse_get_sample_rate_from_index (guint sr_idx)
95 static const guint aac_sample_rates[] = { 96000, 88200, 64000, 48000, 44100,
96 32000, 24000, 22050, 16000, 12000, 11025, 8000
99 if (sr_idx < G_N_ELEMENTS (aac_sample_rates))
100 return aac_sample_rates[sr_idx];
101 GST_WARNING ("Invalid sample rate index %u", sr_idx);
106 * gst_aac_parse_base_init:
107 * @klass: #GstElementClass.
111 gst_aac_parse_base_init (gpointer klass)
113 GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
115 gst_element_class_add_static_pad_template (element_class,
117 gst_element_class_add_static_pad_template (element_class, &src_template);
119 gst_element_class_set_details_simple (element_class,
120 "AAC audio stream parser", "Codec/Parser/Audio",
121 "Advanced Audio Coding parser", "Stefan Kost <stefan.kost@nokia.com>");
126 * gst_aac_parse_class_init:
127 * @klass: #GstAacParseClass.
131 gst_aac_parse_class_init (GstAacParseClass * klass)
133 GstBaseParseClass *parse_class = GST_BASE_PARSE_CLASS (klass);
135 parse_class->start = GST_DEBUG_FUNCPTR (gst_aac_parse_start);
136 parse_class->stop = GST_DEBUG_FUNCPTR (gst_aac_parse_stop);
137 parse_class->set_sink_caps = GST_DEBUG_FUNCPTR (gst_aac_parse_sink_setcaps);
138 parse_class->get_sink_caps = GST_DEBUG_FUNCPTR (gst_aac_parse_sink_getcaps);
139 parse_class->parse_frame = GST_DEBUG_FUNCPTR (gst_aac_parse_parse_frame);
140 parse_class->check_valid_frame =
141 GST_DEBUG_FUNCPTR (gst_aac_parse_check_valid_frame);
146 * gst_aac_parse_init:
147 * @aacparse: #GstAacParse.
148 * @klass: #GstAacParseClass.
152 gst_aac_parse_init (GstAacParse * aacparse, GstAacParseClass * klass)
154 GST_DEBUG ("initialized");
159 * gst_aac_parse_set_src_caps:
160 * @aacparse: #GstAacParse.
161 * @sink_caps: (proposed) caps of sink pad
163 * Set source pad caps according to current knowledge about the
166 * Returns: TRUE if caps were successfully set.
169 gst_aac_parse_set_src_caps (GstAacParse * aacparse, GstCaps * sink_caps)
172 GstCaps *src_caps = NULL;
173 gboolean res = FALSE;
174 const gchar *stream_format;
176 GST_DEBUG_OBJECT (aacparse, "sink caps: %" GST_PTR_FORMAT, sink_caps);
178 src_caps = gst_caps_copy (sink_caps);
180 src_caps = gst_caps_new_simple ("audio/mpeg", NULL);
182 gst_caps_set_simple (src_caps, "framed", G_TYPE_BOOLEAN, TRUE,
183 "mpegversion", G_TYPE_INT, aacparse->mpegversion, NULL);
185 switch (aacparse->header_type) {
186 case DSPAAC_HEADER_NONE:
187 stream_format = "raw";
189 case DSPAAC_HEADER_ADTS:
190 stream_format = "adts";
192 case DSPAAC_HEADER_ADIF:
193 stream_format = "adif";
196 stream_format = NULL;
199 s = gst_caps_get_structure (src_caps, 0);
200 if (aacparse->sample_rate > 0)
201 gst_structure_set (s, "rate", G_TYPE_INT, aacparse->sample_rate, NULL);
202 if (aacparse->channels > 0)
203 gst_structure_set (s, "channels", G_TYPE_INT, aacparse->channels, NULL);
205 gst_structure_set (s, "stream-format", G_TYPE_STRING, stream_format, NULL);
207 GST_DEBUG_OBJECT (aacparse, "setting src caps: %" GST_PTR_FORMAT, src_caps);
209 res = gst_pad_set_caps (GST_BASE_PARSE (aacparse)->srcpad, src_caps);
210 gst_caps_unref (src_caps);
216 * gst_aac_parse_sink_setcaps:
220 * Implementation of "set_sink_caps" vmethod in #GstBaseParse class.
222 * Returns: TRUE on success.
225 gst_aac_parse_sink_setcaps (GstBaseParse * parse, GstCaps * caps)
227 GstAacParse *aacparse;
228 GstStructure *structure;
232 aacparse = GST_AAC_PARSE (parse);
233 structure = gst_caps_get_structure (caps, 0);
234 caps_str = gst_caps_to_string (caps);
236 GST_DEBUG_OBJECT (aacparse, "setcaps: %s", caps_str);
239 /* This is needed at least in case of RTP
240 * Parses the codec_data information to get ObjectType,
241 * number of channels and samplerate */
242 value = gst_structure_get_value (structure, "codec_data");
244 GstBuffer *buf = gst_value_get_buffer (value);
247 const guint8 *buffer = GST_BUFFER_DATA (buf);
250 sr_idx = ((buffer[0] & 0x07) << 1) | ((buffer[1] & 0x80) >> 7);
251 aacparse->object_type = (buffer[0] & 0xf8) >> 3;
252 aacparse->sample_rate = gst_aac_parse_get_sample_rate_from_index (sr_idx);
253 aacparse->channels = (buffer[1] & 0x78) >> 3;
254 aacparse->header_type = DSPAAC_HEADER_NONE;
255 aacparse->mpegversion = 4;
256 aacparse->frame_samples = (buffer[1] & 4) ? 960 : 1024;
258 GST_DEBUG ("codec_data: object_type=%d, sample_rate=%d, channels=%d, "
259 "samples=%d", aacparse->object_type, aacparse->sample_rate,
260 aacparse->channels, aacparse->frame_samples);
262 /* arrange for metadata and get out of the way */
263 gst_aac_parse_set_src_caps (aacparse, caps);
264 gst_base_parse_set_passthrough (parse, TRUE);
268 /* caps info overrides */
269 gst_structure_get_int (structure, "rate", &aacparse->sample_rate);
270 gst_structure_get_int (structure, "channels", &aacparse->channels);
272 gst_base_parse_set_passthrough (parse, FALSE);
280 * gst_aac_parse_adts_get_frame_len:
281 * @data: block of data containing an ADTS header.
283 * This function calculates ADTS frame length from the given header.
285 * Returns: size of the ADTS frame.
288 gst_aac_parse_adts_get_frame_len (const guint8 * data)
290 return ((data[3] & 0x03) << 11) | (data[4] << 3) | ((data[5] & 0xe0) >> 5);
295 * gst_aac_parse_check_adts_frame:
296 * @aacparse: #GstAacParse.
297 * @data: Data to be checked.
298 * @avail: Amount of data passed.
299 * @framesize: If valid ADTS frame was found, this will be set to tell the
300 * found frame size in bytes.
301 * @needed_data: If frame was not found, this may be set to tell how much
302 * more data is needed in the next round to detect the frame
303 * reliably. This may happen when a frame header candidate
304 * is found but it cannot be guaranteed to be the header without
305 * peeking the following data.
307 * Check if the given data contains contains ADTS frame. The algorithm
308 * will examine ADTS frame header and calculate the frame size. Also, another
309 * consecutive ADTS frame header need to be present after the found frame.
310 * Otherwise the data is not considered as a valid ADTS frame. However, this
311 * "extra check" is omitted when EOS has been received. In this case it is
312 * enough when data[0] contains a valid ADTS header.
314 * This function may set the #needed_data to indicate that a possible frame
315 * candidate has been found, but more data (#needed_data bytes) is needed to
316 * be absolutely sure. When this situation occurs, FALSE will be returned.
318 * When a valid frame is detected, this function will use
319 * gst_base_parse_set_min_frame_size() function from #GstBaseParse class
320 * to set the needed bytes for next frame.This way next data chunk is already
323 * Returns: TRUE if the given data contains a valid ADTS header.
326 gst_aac_parse_check_adts_frame (GstAacParse * aacparse,
327 const guint8 * data, const guint avail, gboolean drain,
328 guint * framesize, guint * needed_data)
330 if (G_UNLIKELY (avail < 2))
333 if ((data[0] == 0xff) && ((data[1] & 0xf6) == 0xf0)) {
334 *framesize = gst_aac_parse_adts_get_frame_len (data);
336 /* In EOS mode this is enough. No need to examine the data further.
337 We also relax the check when we have sync, on the assumption that
338 if we're not looking at random data, we have a much higher chance
339 to get the correct sync, and this avoids losing two frames when
340 a single bit corruption happens. */
341 if (drain || !GST_BASE_PARSE_LOST_SYNC (aacparse)) {
345 if (*framesize + ADTS_MAX_SIZE > avail) {
346 /* We have found a possible frame header candidate, but can't be
347 sure since we don't have enough data to check the next frame */
348 GST_DEBUG ("NEED MORE DATA: we need %d, available %d",
349 *framesize + ADTS_MAX_SIZE, avail);
350 *needed_data = *framesize + ADTS_MAX_SIZE;
351 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (aacparse),
352 *framesize + ADTS_MAX_SIZE);
356 if ((data[*framesize] == 0xff) && ((data[*framesize + 1] & 0xf6) == 0xf0)) {
357 guint nextlen = gst_aac_parse_adts_get_frame_len (data + (*framesize));
359 GST_LOG ("ADTS frame found, len: %d bytes", *framesize);
360 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (aacparse),
361 nextlen + ADTS_MAX_SIZE);
368 /* caller ensure sufficient data */
370 gst_aac_parse_parse_adts_header (GstAacParse * aacparse, const guint8 * data,
371 gint * rate, gint * channels, gint * object, gint * version)
375 gint sr_idx = (data[2] & 0x3c) >> 2;
377 *rate = gst_aac_parse_get_sample_rate_from_index (sr_idx);
380 *channels = ((data[2] & 0x01) << 2) | ((data[3] & 0xc0) >> 6);
383 *version = (data[1] & 0x08) ? 2 : 4;
385 *object = (data[2] & 0xc0) >> 6;
389 * gst_aac_parse_detect_stream:
390 * @aacparse: #GstAacParse.
391 * @data: A block of data that needs to be examined for stream characteristics.
392 * @avail: Size of the given datablock.
393 * @framesize: If valid stream was found, this will be set to tell the
394 * first frame size in bytes.
395 * @skipsize: If valid stream was found, this will be set to tell the first
396 * audio frame position within the given data.
398 * Examines the given piece of data and try to detect the format of it. It
399 * checks for "ADIF" header (in the beginning of the clip) and ADTS frame
400 * header. If the stream is detected, TRUE will be returned and #framesize
401 * is set to indicate the found frame size. Additionally, #skipsize might
402 * be set to indicate the number of bytes that need to be skipped, a.k.a. the
403 * position of the frame inside given data chunk.
405 * Returns: TRUE on success.
408 gst_aac_parse_detect_stream (GstAacParse * aacparse,
409 const guint8 * data, const guint avail, gboolean drain,
410 guint * framesize, gint * skipsize)
412 gboolean found = FALSE;
416 GST_DEBUG_OBJECT (aacparse, "Parsing header data");
418 /* FIXME: No need to check for ADIF if we are not in the beginning of the
421 /* Can we even parse the header? */
422 if (avail < ADTS_MAX_SIZE)
425 for (i = 0; i < avail - 4; i++) {
426 if (((data[i] == 0xff) && ((data[i + 1] & 0xf6) == 0xf0)) ||
427 strncmp ((char *) data + i, "ADIF", 4) == 0) {
431 /* Trick: tell the parent class that we didn't find the frame yet,
432 but make it skip 'i' amount of bytes. Next time we arrive
433 here we have full frame in the beginning of the data. */
446 if (gst_aac_parse_check_adts_frame (aacparse, data, avail, drain,
447 framesize, &need_data)) {
450 GST_INFO ("ADTS ID: %d, framesize: %d", (data[1] & 0x08) >> 3, *framesize);
452 aacparse->header_type = DSPAAC_HEADER_ADTS;
453 gst_aac_parse_parse_adts_header (aacparse, data, &rate, &channels,
454 &aacparse->object_type, &aacparse->mpegversion);
456 gst_base_parse_set_frame_rate (GST_BASE_PARSE (aacparse), rate,
457 aacparse->frame_samples, 2, 2);
459 GST_DEBUG ("ADTS: samplerate %d, channels %d, objtype %d, version %d",
460 rate, channels, aacparse->object_type, aacparse->mpegversion);
462 gst_base_parse_set_syncable (GST_BASE_PARSE (aacparse), TRUE);
465 } else if (need_data) {
466 /* This tells the parent class not to skip any data */
471 if (avail < ADIF_MAX_SIZE)
474 if (memcmp (data + i, "ADIF", 4) == 0) {
480 aacparse->header_type = DSPAAC_HEADER_ADIF;
481 aacparse->mpegversion = 4;
483 /* Skip the "ADIF" bytes */
486 /* copyright string */
488 skip_size += 9; /* skip 9 bytes */
490 bitstream_type = adif[0 + skip_size] & 0x10;
492 ((unsigned int) (adif[0 + skip_size] & 0x0f) << 19) |
493 ((unsigned int) adif[1 + skip_size] << 11) |
494 ((unsigned int) adif[2 + skip_size] << 3) |
495 ((unsigned int) adif[3 + skip_size] & 0xe0);
498 if (bitstream_type == 0) {
500 /* Buffer fullness parsing. Currently not needed... */
504 num_elems = (adif[3 + skip_size] & 0x1e);
505 GST_INFO ("ADIF num_config_elems: %d", num_elems);
507 fullness = ((unsigned int) (adif[3 + skip_size] & 0x01) << 19) |
508 ((unsigned int) adif[4 + skip_size] << 11) |
509 ((unsigned int) adif[5 + skip_size] << 3) |
510 ((unsigned int) (adif[6 + skip_size] & 0xe0) >> 5);
512 GST_INFO ("ADIF buffer fullness: %d", fullness);
514 aacparse->object_type = ((adif[6 + skip_size] & 0x01) << 1) |
515 ((adif[7 + skip_size] & 0x80) >> 7);
516 sr_idx = (adif[7 + skip_size] & 0x78) >> 3;
520 aacparse->object_type = (adif[4 + skip_size] & 0x18) >> 3;
521 sr_idx = ((adif[4 + skip_size] & 0x07) << 1) |
522 ((adif[5 + skip_size] & 0x80) >> 7);
525 /* FIXME: This gives totally wrong results. Duration calculation cannot
527 aacparse->sample_rate = gst_aac_parse_get_sample_rate_from_index (sr_idx);
529 /* baseparse is not given any fps,
530 * so it will give up on timestamps, seeking, etc */
532 /* FIXME: Can we assume this? */
533 aacparse->channels = 2;
535 GST_INFO ("ADIF: br=%d, samplerate=%d, objtype=%d",
536 aacparse->bitrate, aacparse->sample_rate, aacparse->object_type);
538 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (aacparse), 512);
540 /* arrange for metadata and get out of the way */
541 gst_aac_parse_set_src_caps (aacparse,
542 GST_PAD_CAPS (GST_BASE_PARSE_SINK_PAD (aacparse)));
544 /* not syncable, not easily seekable (unless we push data from start */
545 gst_base_parse_set_syncable (GST_BASE_PARSE_CAST (aacparse), FALSE);
546 gst_base_parse_set_passthrough (GST_BASE_PARSE_CAST (aacparse), TRUE);
547 gst_base_parse_set_average_bitrate (GST_BASE_PARSE_CAST (aacparse), 0);
553 /* This should never happen */
559 * gst_aac_parse_check_valid_frame:
560 * @parse: #GstBaseParse.
561 * @buffer: #GstBuffer.
562 * @framesize: If the buffer contains a valid frame, its size will be put here
563 * @skipsize: How much data parent class should skip in order to find the
566 * Implementation of "check_valid_frame" vmethod in #GstBaseParse class.
568 * Returns: TRUE if buffer contains a valid frame.
571 gst_aac_parse_check_valid_frame (GstBaseParse * parse,
572 GstBaseParseFrame * frame, guint * framesize, gint * skipsize)
575 GstAacParse *aacparse;
576 gboolean ret = FALSE;
580 aacparse = GST_AAC_PARSE (parse);
581 buffer = frame->buffer;
582 data = GST_BUFFER_DATA (buffer);
584 lost_sync = GST_BASE_PARSE_LOST_SYNC (parse);
586 if (aacparse->header_type == DSPAAC_HEADER_ADIF ||
587 aacparse->header_type == DSPAAC_HEADER_NONE) {
588 /* There is nothing to parse */
589 *framesize = GST_BUFFER_SIZE (buffer);
592 } else if (aacparse->header_type == DSPAAC_HEADER_NOT_PARSED || lost_sync) {
594 ret = gst_aac_parse_detect_stream (aacparse, data, GST_BUFFER_SIZE (buffer),
595 GST_BASE_PARSE_DRAINING (parse), framesize, skipsize);
597 } else if (aacparse->header_type == DSPAAC_HEADER_ADTS) {
598 guint needed_data = 1024;
600 ret = gst_aac_parse_check_adts_frame (aacparse, data,
601 GST_BUFFER_SIZE (buffer), GST_BASE_PARSE_DRAINING (parse),
602 framesize, &needed_data);
605 GST_DEBUG ("buffer didn't contain valid frame");
606 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (aacparse),
611 GST_DEBUG ("buffer didn't contain valid frame");
612 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (aacparse),
621 * gst_aac_parse_parse_frame:
622 * @parse: #GstBaseParse.
623 * @buffer: #GstBuffer.
625 * Implementation of "parse_frame" vmethod in #GstBaseParse class.
627 * Also determines frame overhead.
628 * ADTS streams have a 7 byte header in each frame. MP4 and ADIF streams don't have
629 * a per-frame header.
631 * We're making a couple of simplifying assumptions:
633 * 1. We count Program Configuration Elements rather than searching for them
634 * in the streams to discount them - the overhead is negligible.
636 * 2. We ignore CRC. This has a worst-case impact of (num_raw_blocks + 1)*16
637 * bits, which should still not be significant enough to warrant the
638 * additional parsing through the headers
640 * Returns: GST_FLOW_OK if frame was successfully parsed and can be pushed
641 * forward. Otherwise appropriate error is returned.
644 gst_aac_parse_parse_frame (GstBaseParse * parse, GstBaseParseFrame * frame)
646 GstAacParse *aacparse;
648 GstFlowReturn ret = GST_FLOW_OK;
651 aacparse = GST_AAC_PARSE (parse);
652 buffer = frame->buffer;
654 if (G_UNLIKELY (aacparse->header_type != DSPAAC_HEADER_ADTS))
660 gst_aac_parse_parse_adts_header (aacparse, GST_BUFFER_DATA (buffer),
661 &rate, &channels, NULL, NULL);
662 GST_LOG_OBJECT (aacparse, "rate: %d, chans: %d", rate, channels);
664 if (G_UNLIKELY (rate != aacparse->sample_rate
665 || channels != aacparse->channels)) {
666 aacparse->sample_rate = rate;
667 aacparse->channels = channels;
669 if (!gst_aac_parse_set_src_caps (aacparse,
670 GST_PAD_CAPS (GST_BASE_PARSE (aacparse)->sinkpad))) {
671 /* If linking fails, we need to return appropriate error */
672 ret = GST_FLOW_NOT_LINKED;
675 gst_base_parse_set_frame_rate (GST_BASE_PARSE (aacparse),
676 aacparse->sample_rate, aacparse->frame_samples, 2, 2);
684 * gst_aac_parse_start:
685 * @parse: #GstBaseParse.
687 * Implementation of "start" vmethod in #GstBaseParse class.
689 * Returns: TRUE if startup succeeded.
692 gst_aac_parse_start (GstBaseParse * parse)
694 GstAacParse *aacparse;
696 aacparse = GST_AAC_PARSE (parse);
698 aacparse->frame_samples = 1024;
699 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (aacparse), ADTS_MAX_SIZE);
705 * gst_aac_parse_stop:
706 * @parse: #GstBaseParse.
708 * Implementation of "stop" vmethod in #GstBaseParse class.
710 * Returns: TRUE is stopping succeeded.
713 gst_aac_parse_stop (GstBaseParse * parse)
720 gst_aac_parse_sink_getcaps (GstBaseParse * parse)
725 peercaps = gst_pad_get_allowed_caps (GST_BASE_PARSE_SRC_PAD (parse));
729 /* Remove the framed field */
730 peercaps = gst_caps_make_writable (peercaps);
731 n = gst_caps_get_size (peercaps);
732 for (i = 0; i < n; i++) {
733 GstStructure *s = gst_caps_get_structure (peercaps, i);
735 gst_structure_remove_field (s, "framed");
739 gst_caps_intersect_full (peercaps,
740 gst_pad_get_pad_template_caps (GST_BASE_PARSE_SRC_PAD (parse)),
741 GST_CAPS_INTERSECT_FIRST);
742 gst_caps_unref (peercaps);
745 gst_caps_copy (gst_pad_get_pad_template_caps (GST_BASE_PARSE_SINK_PAD