Fix MediaSource code so it can handle HE-AAC content that uses implicit signalling.

BUG=135674
TEST=Manual testing with the content in the bug.

Review URL: https://chromiumcodereview.appspot.com/10795050

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@147747 0039d316-1c4b-4281-b951-d872f2087c98

6 files changed, 61 insertions, 22 deletions

diff --git a/media/mp4/aac.cc b/media/mp4/aac.cc
index 01c831d..2804d6b 100644
--- a/media/mp4/aac.cc
+++ b/media/mp4/aac.cc
@@ -4,13 +4,15 @@
 
 #include "media/mp4/aac.h"
 
+#include <algorithm>
+
 #include "base/logging.h"
 #include "media/base/bit_reader.h"
 #include "media/mp4/rcheck.h"
 
 // The following conversion table is extracted from ISO 14496 Part 3 -
 // Table 1.16 - Sampling Frequency Index.
-static const uint32 kFrequencyMap[] = {
+static const int kFrequencyMap[] = {
   96000, 88200, 64000, 48000, 44100, 32000, 24000,
   22050, 16000, 12000, 11025, 8000, 7350
 };
@@ -44,7 +46,7 @@ namespace mp4 {
 
 AAC::AAC()
     : profile_(0), frequency_index_(0), channel_config_(0), frequency_(0),
-      channel_layout_(CHANNEL_LAYOUT_UNSUPPORTED) {
+      extension_frequency_(0), channel_layout_(CHANNEL_LAYOUT_UNSUPPORTED) {
 }
 
 AAC::~AAC() {
@@ -57,9 +59,10 @@ bool AAC::Parse(const std::vector<uint8>& data) {
   BitReader reader(&data[0], data.size());
   uint8 extension_type = 0;
   bool ps_present = false;
-  uint8 extension_frequency_index;
+  uint8 extension_frequency_index = 0xff;
 
   frequency_ = 0;
+  extension_frequency_ = 0;
 
   // The following code is written according to ISO 14496 Part 3 Table 1.13 -
   // Syntax of AudioSpecificConfig.
@@ -71,15 +74,13 @@ bool AAC::Parse(const std::vector<uint8>& data) {
     RCHECK(reader.ReadBits(24, &frequency_));
   RCHECK(reader.ReadBits(4, &channel_config_));
 
-  extension_frequency_index = frequency_index_;
-
-  // Read extension configuration
+  // Read extension configuration.
   if (profile_ == 5 || profile_ == 29) {
     ps_present = (profile_ == 29);
     extension_type = 5;
     RCHECK(reader.ReadBits(4, &extension_frequency_index));
     if (extension_frequency_index == 0xf)
-      RCHECK(reader.ReadBits(24, &frequency_));
+      RCHECK(reader.ReadBits(24, &extension_frequency_));
     RCHECK(reader.ReadBits(5, &profile_));
   }
 
@@ -101,7 +102,7 @@ bool AAC::Parse(const std::vector<uint8>& data) {
           RCHECK(reader.ReadBits(4, &extension_frequency_index));
 
           if (extension_frequency_index == 0xf)
-            RCHECK(reader.ReadBits(24, &frequency_));
+            RCHECK(reader.ReadBits(24, &extension_frequency_));
 
           RCHECK(reader.ReadBits(11, &sync_extension_type));
 
@@ -115,8 +116,13 @@ bool AAC::Parse(const std::vector<uint8>& data) {
   }
 
   if (frequency_ == 0) {
+    RCHECK(frequency_index_ < arraysize(kFrequencyMap));
+    frequency_ = kFrequencyMap[frequency_index_];
+  }
+
+  if (extension_frequency_ == 0 && extension_frequency_index != 0xff) {
     RCHECK(extension_frequency_index < arraysize(kFrequencyMap));
-    frequency_ = kFrequencyMap[extension_frequency_index];
+    extension_frequency_ = kFrequencyMap[extension_frequency_index];
   }
 
   // When Parametric Stereo is on, mono will be played as stereo.
@@ -130,8 +136,19 @@ bool AAC::Parse(const std::vector<uint8>& data) {
       channel_config_ <= 7;
 }
 
-uint32 AAC::frequency() const {
-  return frequency_;
+int AAC::GetOutputSamplesPerSecond(bool sbr_in_mimetype) const {
+  if (extension_frequency_ > 0)
+    return extension_frequency_;
+
+  if (!sbr_in_mimetype)
+    return frequency_;
+
+  // The following code is written according to ISO 14496 Part 3 Table 1.11 and
+  // Table 1.22. (Table 1.11 refers to the capping to 48000, Table 1.22 refers
+  // to SBR doubling the AAC sample rate.)
+  // TODO(acolwell) : Extend sample rate cap to 96kHz for Level 5 content.
+  DCHECK_GT(frequency_, 0);
+  return std::min(2 * frequency_, 48000);
 }
 
 ChannelLayout AAC::channel_layout() const {
diff --git a/media/mp4/aac.h b/media/mp4/aac.h
index 7a2a3f8..73464ee 100644
--- a/media/mp4/aac.h
+++ b/media/mp4/aac.h
@@ -32,7 +32,12 @@ class MEDIA_EXPORT AAC {
   // configurations.
   bool Parse(const std::vector<uint8>& data);
 
-  uint32 frequency() const;
+  // Gets the output sample rate for the AAC stream.
+  // |sbr_in_mimetype| should be set to true if the SBR mode is
+  // signalled in the mimetype. (ie mp4a.40.5 in the codecs parameter).
+  // Returns the samples_per_second value that should used in an
+  // AudioDecoderConfig.
+  int GetOutputSamplesPerSecond(bool sbr_in_mimetype) const;
   ChannelLayout channel_layout() const;
 
   // This function converts a raw AAC frame into an AAC frame with an ADTS
@@ -56,7 +61,8 @@ class MEDIA_EXPORT AAC {
   // can be used by Chromium. They are based on the AAC specific
   // configuration but can be overridden by extensions in elementary
   // stream descriptor.
-  uint32 frequency_;
+  int frequency_;
+  int extension_frequency_;
   ChannelLayout channel_layout_;
 };
 
diff --git a/media/mp4/aac_unittest.cc b/media/mp4/aac_unittest.cc
index 2f61de5..c9f888c 100644
--- a/media/mp4/aac_unittest.cc
+++ b/media/mp4/aac_unittest.cc
@@ -18,7 +18,7 @@ TEST(AACTest, BasicProfileTest) {
   data.assign(buffer, buffer + sizeof(buffer));
 
   EXPECT_TRUE(aac.Parse(data));
-  EXPECT_EQ(aac.frequency(), 44100u);
+  EXPECT_EQ(aac.GetOutputSamplesPerSecond(false), 44100);
   EXPECT_EQ(aac.channel_layout(), CHANNEL_LAYOUT_STEREO);
 }
 
@@ -30,7 +30,21 @@ TEST(AACTest, ExtensionTest) {
   data.assign(buffer, buffer + sizeof(buffer));
 
   EXPECT_TRUE(aac.Parse(data));
-  EXPECT_EQ(aac.frequency(), 48000u);
+  EXPECT_EQ(aac.GetOutputSamplesPerSecond(false), 48000);
+  EXPECT_EQ(aac.GetOutputSamplesPerSecond(true), 48000);
+  EXPECT_EQ(aac.channel_layout(), CHANNEL_LAYOUT_STEREO);
+}
+
+TEST(AACTest, TestImplicitSBR) {
+  AAC aac;
+  uint8 buffer[] = {0x13, 0x10};
+  std::vector<uint8> data;
+
+  data.assign(buffer, buffer + sizeof(buffer));
+
+  EXPECT_TRUE(aac.Parse(data));
+  EXPECT_EQ(aac.GetOutputSamplesPerSecond(false), 24000);
+  EXPECT_EQ(aac.GetOutputSamplesPerSecond(true), 48000);
   EXPECT_EQ(aac.channel_layout(), CHANNEL_LAYOUT_STEREO);
 }
 
@@ -42,7 +56,7 @@ TEST(AACTest, SixChannelTest) {
   data.assign(buffer, buffer + sizeof(buffer));
 
   EXPECT_TRUE(aac.Parse(data));
-  EXPECT_EQ(aac.frequency(), 48000u);
+  EXPECT_EQ(aac.GetOutputSamplesPerSecond(false), 48000);
   EXPECT_EQ(aac.channel_layout(), CHANNEL_LAYOUT_5_1);
 }
 
diff --git a/media/mp4/mp4_stream_parser.cc b/media/mp4/mp4_stream_parser.cc
index 45d81db..05b18c1 100644
--- a/media/mp4/mp4_stream_parser.cc
+++ b/media/mp4/mp4_stream_parser.cc
@@ -19,14 +19,15 @@
 namespace media {
 namespace mp4 {
 
-MP4StreamParser::MP4StreamParser()
+MP4StreamParser::MP4StreamParser(bool has_sbr)
     : state_(kWaitingForInit),
       moof_head_(0),
       mdat_tail_(0),
       has_audio_(false),
       has_video_(false),
       audio_track_id_(0),
-      video_track_id_(0) {
+      video_track_id_(0),
+      has_sbr_(has_sbr) {
 }
 
 MP4StreamParser::~MP4StreamParser() {}
@@ -187,9 +188,9 @@ bool MP4StreamParser::ParseMoov(BoxReader* reader) {
       // Check if it is MPEG4 AAC defined in ISO 14496 Part 3.
       RCHECK(entry.esds.object_type == kISO_14496_3);
       audio_config.Initialize(kCodecAAC, entry.samplesize,
-                              aac.channel_layout(), aac.frequency(),
+                              aac.channel_layout(),
+                              aac.GetOutputSamplesPerSecond(has_sbr_),
                               NULL, 0, false);
-
       has_audio_ = true;
       audio_track_id_ = track->header.track_id;
     }
diff --git a/media/mp4/mp4_stream_parser.h b/media/mp4/mp4_stream_parser.h
index a830536..a19bbcc 100644
--- a/media/mp4/mp4_stream_parser.h
+++ b/media/mp4/mp4_stream_parser.h
@@ -22,7 +22,7 @@ class BoxReader;
 
 class MEDIA_EXPORT MP4StreamParser : public StreamParser {
  public:
-  MP4StreamParser();
+  MP4StreamParser(bool has_sbr);
   virtual ~MP4StreamParser();
 
   virtual void Init(const InitCB& init_cb, const NewConfigCB& config_cb,
@@ -83,6 +83,7 @@ class MEDIA_EXPORT MP4StreamParser : public StreamParser {
   bool has_video_;
   uint32 audio_track_id_;
   uint32 video_track_id_;
+  bool has_sbr_;
 
   DISALLOW_COPY_AND_ASSIGN(MP4StreamParser);
 };
diff --git a/media/mp4/mp4_stream_parser_unittest.cc b/media/mp4/mp4_stream_parser_unittest.cc
index f5be02af..db2555f 100644
--- a/media/mp4/mp4_stream_parser_unittest.cc
+++ b/media/mp4/mp4_stream_parser_unittest.cc
@@ -27,7 +27,7 @@ namespace mp4 {
 class MP4StreamParserTest : public testing::Test {
  public:
   MP4StreamParserTest()
-      : parser_(new MP4StreamParser),
+      : parser_(new MP4StreamParser(false)),
         got_configs_(false) {
   }