Mpeg2 TS stream parser for media source.

media/mp2t/es_parser_adts.cc

+// Copyright (c) 2013 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "media/mp2t/es_parser_adts.h"
+
+#include <list>
+
+#include "base/basictypes.h"
+#include "base/logging.h"
+#include "base/strings/string_number_conversions.h"
+#include "media/base/audio_decoder_config.h"

[acolwell GONE FROM CHROMIUM, 2013/09/16 06:19:30]

nit: Shouldn't need this now since it is in the header.

[damienv1, 2013/09/17 02:58:22]

Done.

+#include "media/base/audio_timestamp_helper.h"
+#include "media/base/bit_reader.h"
+#include "media/base/buffers.h"
+#include "media/base/channel_layout.h"
+#include "media/base/stream_parser_buffer.h"
+#include "media/mp2t/mp2t_common.h"
+
+// Adts header is at least 7 bytes (can be 9 bytes).
+static const int kAdtsHeaderMinSize = 7;
+
+static const int adts_frequency_table[16] = {
+  96000,
+  88200,
+  64000,
+  48000,
+  44100,
+  32000,
+  24000,
+  22050,
+  16000,
+  12000,
+  11025,
+  8000,
+  7350,
+  0,
+  0,
+  0,
+};
+static const int kMaxSupportedFrequencyIndex = 12;
+
+static media::ChannelLayout adts_channel_layout[8] = {
+  media::CHANNEL_LAYOUT_NONE,
+  media::CHANNEL_LAYOUT_MONO,
+  media::CHANNEL_LAYOUT_STEREO,
+  media::CHANNEL_LAYOUT_SURROUND,
+  media::CHANNEL_LAYOUT_4_0,
+  media::CHANNEL_LAYOUT_5_0_BACK,
+  media::CHANNEL_LAYOUT_5_1_BACK,
+  media::CHANNEL_LAYOUT_7_1,
+};
+
+// Number of samples per frame.
+static const int kNumberSamplesPerAACFrame = 1024;
+
+static int ExtractAdtsFrameSize(const uint8* adts_header) {
+  return ((static_cast<int>(adts_header[5]) >> 5) |
+          (static_cast<int>(adts_header[4]) << 3) |
+          ((static_cast<int>(adts_header[3]) & 0x3) << 11));
+}
+
+static int ExtractAdtsFrequencyIndex(const uint8* adts_header) {
+  return ((adts_header[2] >> 2) & 0xf);
+}
+
+static int ExtractAdtsChannelConfig(const uint8* adts_header) {
+  return (((adts_header[3] >> 6) & 0x3) |
+          ((adts_header[2] & 0x1) << 2));
+}
+
+// Look for an ADTS syncword.
+// |new_pos| returns
+// - either the byte position of the ADTS frame (if found)
+// - or the byte position of 1st byte that was not processed (if not found).
+// |frame_sz| returns the size of the ADTS frame (if found).
+// Return whether a syncword was found.
+static bool LookForSyncWord(const uint8* raw_es, int raw_es_size,
+                            int pos,
+                            int* new_pos, int* frame_sz) {
+  int max_offset = raw_es_size - kAdtsHeaderMinSize;

[acolwell GONE FROM CHROMIUM, 2013/09/16 06:19:30]

nit: 
DCHECK_GE(pos, 0);
DCHECK_LT(pos, raw_es_size);

[damienv1, 2013/09/17 02:58:22]

Done.
Except that this should be DCHECK_LE(pos, raw_es_size);
As |pos| can point to the 1st byte that was not processed yet.
So if the last byte of the buffer was processed, pos is pointing to
|raw_es_size|.

+  if (max_offset < 0)
+    max_offset = 0;

[acolwell GONE FROM CHROMIUM, 2013/09/16 06:19:30]

nit: I think it might be clearer if you just set *new_pos and early return here.

Also I think there might be a bug in the existing code if pos > 0 && max_offset
< 0. *new_pos should always be == to pos in this case right?

[damienv1, 2013/09/17 02:58:22]

Yes, the new position should remain unchanged if max_offset < 0.
I updated the behavior.

+
+  for (int offset = pos; offset < max_offset; offset++) {
+    const uint8* cur_buf = &raw_es[offset];
+
+    if ((cur_buf[0] != 0xff) || ((cur_buf[1] & 0xf6) != 0xf0))
+      // The first 12 bits must be 1.
+      // The layer field (2 bits) must be set to 0.
+      continue;
+
+    int frame_size = ExtractAdtsFrameSize(cur_buf);
+    if (frame_size < kAdtsHeaderMinSize)
+      // Too short to be an ADTS frame.
+      continue;
+
+    // Check whether there is another frame
+    // |size| apart from the current one.
+    int remaining_size = raw_es_size - offset;
+    if (remaining_size >= frame_size + 2) {
+      if ((cur_buf[frame_size] != 0xff) ||

[acolwell GONE FROM CHROMIUM, 2013/09/16 06:19:30]

nit: Replace this and the check above w/ isSyncWord() helper call just to make
the code a little clearer?

[damienv1, 2013/09/17 02:58:22]

Done.

+          (cur_buf[frame_size + 1] & 0xf6) != 0xf0)
+        continue;
+    }
+
+    *new_pos = offset;
+    *frame_sz = frame_size;
+    return true;
+  }
+
+  *new_pos = max_offset;
+  return false;
+}
+
+namespace media {
+namespace mp2t {
+
+EsParserAdts::EsParserAdts(
+    NewAudioConfigCB new_audio_config_cb,
+    EmitBufferCB emit_buffer_cb)
+  : new_audio_config_cb_(new_audio_config_cb),
+    emit_buffer_cb_(emit_buffer_cb) {
+}
+
+EsParserAdts::~EsParserAdts() {
+}
+
+bool EsParserAdts::Parse(const uint8* buf, int size,
+                         base::TimeDelta pts,
+                         base::TimeDelta dts) {
+  int raw_es_size = 0;
+  const uint8* raw_es = NULL;

[acolwell GONE FROM CHROMIUM, 2013/09/16 06:19:30]

nit: var init isn't needed here.

[damienv1, 2013/09/17 02:58:22]

Done.

+
+  // The incoming PTS applies to the access unit that comes just after
+  // the beginning of |buf|.
+  if (pts != kNoTimestamp()) {
+    es_byte_queue_.Peek(&raw_es, &raw_es_size);
+    pts_list_.push_back(EsPts(raw_es_size, pts));
+  }
+
+  // Copy the input data to the ES buffer.
+  es_byte_queue_.Push(buf, size);
+  es_byte_queue_.Peek(&raw_es, &raw_es_size);
+
+  // Look for every ADTS frame in the ES buffer starting at offset = 0
+  int es_position = 0;
+  int frame_size;
+  while (LookForSyncWord(raw_es, raw_es_size, es_position,
+                         &es_position, &frame_size)) {
+    DVLOG(LOG_LEVEL_ES) << "ADTS syncword @ pos=" << es_position
+                        << " frame_size=" << frame_size;
+    DVLOG(LOG_LEVEL_ES) << "ADTS header: "
+                        << base::HexEncode(&raw_es[es_position], 7);
+
+    // Do not process the frame if this one is a partial frame.
+    int remaining_size = raw_es_size - es_position;
+    if (frame_size > remaining_size)
+      break;
+
+    // Update the audio configuration if needed.
+    DCHECK_GE(frame_size, kAdtsHeaderMinSize);
+    if (!UpdateAudioConfiguration(&raw_es[es_position]))
+      return false;
+
+    // Get the PTS & the duration of this access unit.
+    DCHECK(audio_timestamp_helper_);

[acolwell GONE FROM CHROMIUM, 2013/09/16 06:19:30]

nit: No need for DCHECK here since the code will crash in 4 or 7 lines later.
Usually in these types of situations we just let the code crash instead of
putting a DCHECK. We tend to use NULL DCHECKs only for method param
verification.

[damienv1, 2013/09/17 02:58:22]

Done.

+    while (!pts_list_.empty() &&
+           pts_list_.front().first <= es_position) {
+      audio_timestamp_helper_->SetBaseTimestamp(pts_list_.front().second);

[acolwell GONE FROM CHROMIUM, 2013/09/16 06:19:30]

Is this here to compensate for gaps in the audio? You shouldn't need to do this
otherwise since the accumulation of samples should be more accurate and ensures
that gaps/overlaps aren't created. I think you should only reset the base if the
timestamp is more than a few samples different.

[damienv1, 2013/09/17 02:58:22]

1) The timestamp helper has a resolution of 1 sample. The Mpeg2 TS timestamps
have a resolution of 90kHz which in case of AAC is less than a sample (except
for 96kHz where it's slightly more).

2) The Mpeg2 TS timestamp should be the one to use as there might be some GAP in
the audio as well. I think the ADTS parser should not be the one to decide which
PTS to use when there is one coming from the stream.

+      pts_list_.pop_front();
+    }
+
+    base::TimeDelta current_pts = audio_timestamp_helper_->GetTimestamp();
+    base::TimeDelta frame_duration =
+        audio_timestamp_helper_->GetFrameDuration(kNumberSamplesPerAACFrame);
+
+    // Emit an audio frame.
+    bool is_key_frame = true;
+    scoped_refptr<StreamParserBuffer> stream_parser_buffer =
+        StreamParserBuffer::CopyFrom(
+            &raw_es[es_position],
+            frame_size,
+            is_key_frame);
+    stream_parser_buffer->SetDecodeTimestamp(current_pts);
+    stream_parser_buffer->set_timestamp(current_pts);
+    stream_parser_buffer->set_duration(frame_duration);
+    emit_buffer_cb_.Run(stream_parser_buffer);
+
+    // Update the PTS of the next frame.
+    audio_timestamp_helper_->AddFrames(kNumberSamplesPerAACFrame);
+
+    // Skip the current frame.
+    es_position += frame_size;
+  }
+
+  // Discard all the bytes that have been processed.
+  DiscardEs(es_position);
+
+  return true;
+}
+
+void EsParserAdts::Flush() {
+  // All the complete frames have been emitted,
+  // so just clear the ES buffer.
+  es_byte_queue_.Reset();
+  pts_list_.clear();

[acolwell GONE FROM CHROMIUM, 2013/09/16 06:19:30]

This doesn't seem consistent w/ the description in es_parser.h. I'd expect this
function to be a noop, but it looks like it is effecting the parser state. Am I
misunderstanding something here?

[damienv1, 2013/09/17 02:58:22]

I made it a noop to match the description.
It does not impact the behavior since a Flush is always followed by a Reset in
practice.

+}
+
+void EsParserAdts::Reset() {
+  es_byte_queue_.Reset();
+  pts_list_.clear();
+  last_audio_decoder_config_ = AudioDecoderConfig();
+}
+
+bool EsParserAdts::UpdateAudioConfiguration(const uint8* adts_header) {
+  int frequency_index = ExtractAdtsFrequencyIndex(adts_header);
+  if (frequency_index > kMaxSupportedFrequencyIndex) {
+    // Frequency index 13 & 14 are reserved
+    // while 15 means that the frequency is explicitly written
+    // (not supported).
+    return false;
+  }
+
+  // TODO(damienv): support HE-AAC frequency doubling (SBR)
+  // based on the incoming ADTS profile.
+  int samples_per_second = adts_frequency_table[frequency_index];
+  int channel_configuration = ExtractAdtsChannelConfig(adts_header);
+  int adts_profile = (adts_header[2] >> 6) & 0x3;
+

[acolwell GONE FROM CHROMIUM, 2013/09/16 06:19:30]

I think you need the following since I don't believe we support inband-channel
layout.

// TODO: Add support for inband channel configurations.
if (channel_configuration == 0)
  return false;

[damienv1, 2013/09/17 02:58:22]

Done.

+  AudioDecoderConfig audio_decoder_config(
+      kCodecAAC,
+      kSampleFormatS16,
+      adts_channel_layout[channel_configuration],
+      samples_per_second,
+      NULL, 0,
+      false);
+
+  if (!audio_decoder_config.Matches(last_audio_decoder_config_)) {
+    DVLOG(1) << "Sampling frequency: " << samples_per_second;
+    DVLOG(1) << "Channel config: " << channel_configuration;
+    DVLOG(1) << "Adts profile: " << adts_profile;
+    // Reset the timestamp helper to use a new time scale.
+    if (audio_timestamp_helper_) {
+      base::TimeDelta base_timestamp = audio_timestamp_helper_->GetTimestamp();
+      audio_timestamp_helper_.reset(
+        new AudioTimestampHelper(samples_per_second));
+      audio_timestamp_helper_->SetBaseTimestamp(base_timestamp);
+    } else {
+      audio_timestamp_helper_.reset(
+          new AudioTimestampHelper(samples_per_second));
+    }
+    // Audio config notification.
+    last_audio_decoder_config_ = audio_decoder_config;
+    new_audio_config_cb_.Run(audio_decoder_config);
+  }
+
+  return true;
+}
+
+void EsParserAdts::DiscardEs(int nbytes) {
+  DCHECK_GE(nbytes, 0);
+  if (nbytes <= 0)
+    return;
+
+  // Adjust the ES position of each PTS.
+  for (EsPtsList::iterator it = pts_list_.begin(); it != pts_list_.end(); ++it)
+    it->first -= nbytes;

[acolwell GONE FROM CHROMIUM, 2013/09/16 06:19:30]

How big is this list likely to be? It might be more efficient to keep a
bytes_consumed_ count and then just add the offset when you insert the elements
& subract it off when you are popping them off.

[damienv1, 2013/09/17 02:58:22]

In the general case, the size of the list will be 0 at this time or possibly 1
(if there is no full ADTS frame in the last reassembled PES). "List size > 1" is
very unlikely since this would mean:
- an ADTS frame spreads over multiple reassembled PES
- or no ADTS header was found over the last couple of reassembled PES

At any time in this code, the PTS list size should be between 0 and 2 included.
Any other value (> 2) is very unlikely (though not impossible).

+
+  // Discard |nbytes| of ES.
+  es_byte_queue_.Pop(nbytes);
+}
+
+}  // namespace mp2t
+}  // namespace media
+