mpegtsmux: Fix handling of MPEG-2 AAC

The audio/mpeg,mpegversion=2 caps in GStreamer refer to
MPEG-2 AAC (ISO 13818-7), not to the extended MP3 (ISO 13818-3),
which is audio/mpeg,mpegversion=1,mpegaudioversion=2/3

Fix the caps, and add handling for MPEG-2 AAC in both ADTS and raw
form, adding ADTS headers for the latter.
This commit is contained in:
Jan Schmidt 2020-06-18 04:03:59 +10:00 committed by Jan Schmidt
parent 4c4b3dfc76
commit 46cc64e09f
5 changed files with 176 additions and 35 deletions

View file

@ -208846,7 +208846,7 @@
"long-name": "MPEG Transport Stream Muxer",
"pad-templates": {
"sink_%%d": {
"caps": "video/mpeg:\n parsed: true\n mpegversion: { (int)1, (int)2, (int)4 }\n systemstream: false\nvideo/x-dirac:\nimage/x-jpc:\nvideo/x-h264:\n stream-format: byte-stream\n alignment: { (string)au, (string)nal }\nvideo/x-h265:\n stream-format: byte-stream\n alignment: { (string)au, (string)nal }\naudio/mpeg:\n parsed: true\n mpegversion: { (int)1, (int)2 }\naudio/mpeg:\n framed: true\n mpegversion: 4\n stream-format: adts\naudio/mpeg:\n mpegversion: 4\n stream-format: raw\naudio/x-lpcm:\n width: { (int)16, (int)20, (int)24 }\n rate: { (int)48000, (int)96000 }\n channels: [ 1, 8 ]\n dynamic_range: [ 0, 255 ]\n emphasis: { (boolean)false, (boolean)true }\n mute: { (boolean)false, (boolean)true }\naudio/x-ac3:\n framed: true\naudio/x-dts:\n framed: true\naudio/x-opus:\n channels: [ 1, 8 ]\nchannel-mapping-family: { (int)0, (int)1 }\nsubpicture/x-dvb:\napplication/x-teletext:\nmeta/x-klv:\n parsed: true\nimage/x-jpc:\n profile: [ 0, 49151 ]\n",
"caps": "video/mpeg:\n parsed: true\n mpegversion: { (int)1, (int)2, (int)4 }\n systemstream: false\nvideo/x-dirac:\nimage/x-jpc:\nvideo/x-h264:\n stream-format: byte-stream\n alignment: { (string)au, (string)nal }\nvideo/x-h265:\n stream-format: byte-stream\n alignment: { (string)au, (string)nal }\naudio/mpeg:\n parsed: true\n mpegversion: 1\naudio/mpeg:\n framed: true\n mpegversion: { (int)2, (int)4 }\n stream-format: { (string)adts, (string)raw }\naudio/x-lpcm:\n width: { (int)16, (int)20, (int)24 }\n rate: { (int)48000, (int)96000 }\n channels: [ 1, 8 ]\n dynamic_range: [ 0, 255 ]\n emphasis: { (boolean)false, (boolean)true }\n mute: { (boolean)false, (boolean)true }\naudio/x-ac3:\n framed: true\naudio/x-dts:\n framed: true\naudio/x-opus:\n channels: [ 1, 8 ]\nchannel-mapping-family: { (int)0, (int)1 }\nsubpicture/x-dvb:\napplication/x-teletext:\nmeta/x-klv:\n parsed: true\nimage/x-jpc:\n profile: [ 0, 49151 ]\n",
"direction": "sink",
"presence": "request",
"type": "GstBaseTsMuxPad"

View file

@ -389,6 +389,7 @@ gst_base_ts_mux_create_stream (GstBaseTsMux * mux, GstBaseTsMuxPad * ts_pad)
guint32 max_rate = 0;
guint8 color_spec = 0;
j2k_private_data *private_data = NULL;
const gchar *stream_format = NULL;
pad = GST_PAD (ts_pad);
caps = gst_pad_get_current_caps (pad);
@ -405,6 +406,8 @@ gst_base_ts_mux_create_stream (GstBaseTsMux * mux, GstBaseTsMuxPad * ts_pad)
if (value != NULL)
codec_data = gst_value_get_buffer (value);
stream_format = gst_structure_get_string (s, "stream-format");
if (strcmp (mt, "video/x-dirac") == 0) {
st = TSMUX_ST_VIDEO_DIRAC;
} else if (strcmp (mt, "audio/x-ac3") == 0) {
@ -426,23 +429,50 @@ gst_base_ts_mux_create_stream (GstBaseTsMux * mux, GstBaseTsMuxPad * ts_pad)
}
switch (mpegversion) {
case 1:
case 1:{
int mpegaudioversion = 1; /* Assume mpegaudioversion=1 for backwards compatibility */
(void) gst_structure_get_int (s, "mpegaudioversion", &mpegaudioversion);
if (mpegaudioversion == 1)
st = TSMUX_ST_AUDIO_MPEG1;
break;
case 2:
else
st = TSMUX_ST_AUDIO_MPEG2;
break;
}
case 2:{
/* mpegversion=2 in GStreamer refers to MPEG-2 Part 7 audio, */
st = TSMUX_ST_AUDIO_AAC;
/* Check the stream format. If raw, make dummy internal codec data from the caps */
if (g_strcmp0 (stream_format, "raw") == 0) {
ts_pad->codec_data =
gst_base_ts_mux_aac_mpeg2_make_codec_data (mux, caps);
ts_pad->prepare_func = gst_base_ts_mux_prepare_aac_mpeg2;
if (ts_pad->codec_data == NULL) {
GST_ERROR_OBJECT (mux, "Invalid or incomplete caps for MPEG-2 AAC");
goto not_negotiated;
}
}
break;
}
case 4:
{
st = TSMUX_ST_AUDIO_AAC;
if (codec_data) { /* TODO - Check stream format - codec data should only come with RAW stream */
/* Check the stream format. We need codec_data with RAW streams and mpegversion=4 */
if (g_strcmp0 (stream_format, "raw") == 0) {
if (codec_data) {
GST_DEBUG_OBJECT (pad,
"we have additional codec data (%" G_GSIZE_FORMAT " bytes)",
gst_buffer_get_size (codec_data));
ts_pad->codec_data = gst_buffer_ref (codec_data);
ts_pad->prepare_func = gst_base_ts_mux_prepare_aac;
ts_pad->prepare_func = gst_base_ts_mux_prepare_aac_mpeg4;
} else {
ts_pad->codec_data = NULL;
GST_ERROR_OBJECT (mux, "Need codec_data for raw MPEG-4 AAC");
goto not_negotiated;
}
}
break;
}

View file

@ -84,44 +84,40 @@
#include "config.h"
#endif
#include <gst/pbutils/pbutils.h>
#include "gstbasetsmuxaac.h"
#include <string.h>
#define GST_CAT_DEFAULT gst_base_ts_mux_debug
GstBuffer *
gst_base_ts_mux_prepare_aac (GstBuffer * buf, GstBaseTsMuxPad * pad,
GstBaseTsMux * mux)
static GstBuffer *
gst_base_ts_mux_prepare_aac_adts (GstBuffer * buf,
GstBaseTsMux * mux, gboolean is_mpeg2, guint8 obj_type_profile,
guint8 rate_idx, guint8 channels)
{
guint8 adts_header[7] = { 0, };
gsize out_size = gst_buffer_get_size (buf) + 7;
GstBuffer *out_buf = gst_buffer_new_and_alloc (out_size);
gsize out_offset = 0;
guint8 rate_idx = 0, channels = 0, obj_type = 0;
GstMapInfo codec_data_map;
GstMapInfo buf_map;
/* Generate ADTS header */
GST_DEBUG_OBJECT (mux, "Preparing AAC buffer for output");
gst_buffer_copy_into (out_buf, buf,
GST_BUFFER_COPY_METADATA | GST_BUFFER_COPY_TIMESTAMPS, 0, 0);
gst_buffer_map (pad->codec_data, &codec_data_map, GST_MAP_READ);
GST_DEBUG_OBJECT (mux, "Rate index %u, channels %u, object type/profile %u",
rate_idx, channels, obj_type_profile);
/* Generate ADTS header */
obj_type = GST_READ_UINT8 (codec_data_map.data) >> 3;
rate_idx = (GST_READ_UINT8 (codec_data_map.data) & 0x7) << 1;
rate_idx |= (GST_READ_UINT8 (codec_data_map.data + 1) & 0x80) >> 7;
channels = (GST_READ_UINT8 (codec_data_map.data + 1) & 0x78) >> 3;
GST_DEBUG_OBJECT (mux, "Rate index %u, channels %u, object type %u", rate_idx,
channels, obj_type);
/* Sync point over a full byte */
adts_header[0] = 0xFF;
/* Sync point continued over first 4 bits + static 4 bits
* (ID, layer, protection)*/
adts_header[1] = 0xF1;
/* Object type over first 2 bits */
adts_header[2] = (obj_type - 1) << 6;
adts_header[1] = 0xF1 | (is_mpeg2 ? 0x8 : 0x0);
/* Object type (MPEG4) / Profile (MPEG2) over first 2 bits */
adts_header[2] = (obj_type_profile - 1) << 6;
/* rate index over next 4 bits */
adts_header[2] |= (rate_idx << 2);
/* channels over last 2 bits */
@ -149,8 +145,119 @@ gst_base_ts_mux_prepare_aac (GstBuffer * buf, GstBaseTsMuxPad * pad,
/* Now copy complete frame */
gst_buffer_fill (out_buf, out_offset, buf_map.data, buf_map.size);
gst_buffer_unmap (pad->codec_data, &codec_data_map);
gst_buffer_unmap (buf, &buf_map);
return out_buf;
}
/* Constructs a dummy codec_data buffer for generating ADTS headers
* from raw MPEG-2 AAC input, where we don't expect codec_data in the caps,
* and need to get the info from the profile/channels/rate fields */
GstBuffer *
gst_base_ts_mux_aac_mpeg2_make_codec_data (GstBaseTsMux * mux,
const GstCaps * caps)
{
const GstStructure *s;
const gchar *profile_str;
gint channels, rate;
guint8 profile_idx, channel_idx;
gint rate_idx;
GstMapInfo map;
GstBuffer *ret;
s = gst_caps_get_structure (caps, 0);
profile_str = gst_structure_get_string (s, "profile");
if (G_UNLIKELY (profile_str == NULL)) {
GST_ERROR_OBJECT (mux, "AAC caps do not contain profile");
return NULL;
}
if (G_UNLIKELY (!gst_structure_get_int (s, "rate", &rate))) {
GST_ERROR_OBJECT (mux, "AAC caps do not contain a sample rate");
return NULL;
}
if (G_UNLIKELY (!gst_structure_get_int (s, "channels", &channels))) {
GST_ERROR_OBJECT (mux, "AAC caps do not contain channel count");
return NULL;
}
if (g_strcmp0 (profile_str, "main") == 0) {
profile_idx = (guint8) 0U;
} else if (g_strcmp0 (profile_str, "lc") == 0) {
profile_idx = (guint8) 1U;
} else if (g_strcmp0 (profile_str, "ssr") == 0) {
profile_idx = (guint8) 2U;
} else {
GST_ERROR_OBJECT (mux, "Invalid profile %s for MPEG-2 AAC caps",
profile_str);
return NULL;
}
if (channels >= 1 && channels <= 6) /* Mono up to & including 5.1 */
channel_idx = (guint8) channels;
else if (channels == 8) /* 7.1 */
channel_idx = (guint8) 7U;
else {
GST_ERROR_OBJECT (mux, "Invalid channel count %d for MPEG-2 AAC caps",
channels);
return NULL;
}
rate_idx = gst_codec_utils_aac_get_index_from_sample_rate (rate);
if (rate_idx < 0) {
GST_ERROR_OBJECT (mux, "Invalid samplerate %d for MPEG-2 AAC caps", rate);
return NULL;
}
ret = gst_buffer_new_and_alloc (3);
gst_buffer_map (ret, &map, GST_MAP_READ);
map.data[0] = profile_idx;
map.data[1] = (guint8) rate_idx;
map.data[2] = channel_idx;
gst_buffer_unmap (ret, &map);
return ret;
}
GstBuffer *
gst_base_ts_mux_prepare_aac_mpeg4 (GstBuffer * buf, GstBaseTsMuxPad * pad,
GstBaseTsMux * mux)
{
GstMapInfo codec_data_map;
guint8 rate_idx = 0, channels = 0, obj_type = 0;
g_return_val_if_fail (pad->codec_data != NULL, NULL);
gst_buffer_map (pad->codec_data, &codec_data_map, GST_MAP_READ);
obj_type = GST_READ_UINT8 (codec_data_map.data) >> 3;
rate_idx = (GST_READ_UINT8 (codec_data_map.data) & 0x7) << 1;
rate_idx |= (GST_READ_UINT8 (codec_data_map.data + 1) & 0x80) >> 7;
channels = (GST_READ_UINT8 (codec_data_map.data + 1) & 0x78) >> 3;
gst_buffer_unmap (pad->codec_data, &codec_data_map);
return gst_base_ts_mux_prepare_aac_adts (buf, mux, FALSE, obj_type, rate_idx,
channels);
}
GstBuffer *
gst_base_ts_mux_prepare_aac_mpeg2 (GstBuffer * buf, GstBaseTsMuxPad * pad,
GstBaseTsMux * mux)
{
GstMapInfo codec_data_map;
guint8 rate_idx = 0, channels = 0, profile_obj_type = 0;
g_return_val_if_fail (pad->codec_data != NULL, NULL);
/* Dummy codec data with 3 bytes of profile_idx, rate_idx, channel_idx */
gst_buffer_map (pad->codec_data, &codec_data_map, GST_MAP_READ);
profile_obj_type = GST_READ_UINT8 (codec_data_map.data);
rate_idx = GST_READ_UINT8 (codec_data_map.data + 1);
channels = GST_READ_UINT8 (codec_data_map.data + 2);
gst_buffer_unmap (pad->codec_data, &codec_data_map);
return gst_base_ts_mux_prepare_aac_adts (buf, mux, TRUE, profile_obj_type,
rate_idx, channels);
}

View file

@ -85,7 +85,13 @@
#include "gstbasetsmux.h"
GstBuffer * gst_base_ts_mux_prepare_aac (GstBuffer * buf, GstBaseTsMuxPad * pad,
GstBuffer * gst_base_ts_mux_prepare_aac_mpeg4 (GstBuffer * buf, GstBaseTsMuxPad * pad,
GstBaseTsMux * mux);
GstBuffer * gst_base_ts_mux_aac_mpeg2_make_codec_data (GstBaseTsMux * mux, const GstCaps *caps);
GstBuffer * gst_base_ts_mux_prepare_aac_mpeg2 (GstBuffer * buf, GstBaseTsMuxPad * pad,
GstBaseTsMux * mux);
#endif /* __BASETSMUX_AAC_H__ */

View file

@ -122,12 +122,10 @@ static GstStaticPadTemplate gst_mpeg_ts_mux_sink_factory =
"alignment=(string){au, nal}; "
"audio/mpeg, "
"parsed = (boolean) TRUE, "
"mpegversion = (int) { 1, 2 };"
"mpegversion = (int) 1;"
"audio/mpeg, "
"framed = (boolean) TRUE, "
"mpegversion = (int) 4, stream-format = (string) adts;"
"audio/mpeg, "
"mpegversion = (int) 4, stream-format = (string) raw;"
"mpegversion = (int) {2, 4}, stream-format = (string) { adts, raw };"
"audio/x-lpcm, "
"width = (int) { 16, 20, 24 }, "
"rate = (int) { 48000, 96000 }, "