126 lines
4.9 KiB
C++
126 lines
4.9 KiB
C++
/**
|
|
* @file decode-mpg.ino
|
|
* @brief Desktop counterpart of sd-mpg-audio-video.ino: plays a local MPEG-1
|
|
* Program Stream (.mpg) file - video via DemuxerMPG -> MPGDecoder (TinyMPG)
|
|
* -> OutputOpenCV, audio via DecoderHelix (auto-selects MP3/AAC/WAV from
|
|
* DemuxerMPG's mime()) -> PortAudioStream. A dedicated MP2Decoder (TinyMP2)
|
|
* is registered on top for Layer II specifically - see DemuxerMPG::mime(),
|
|
* which reports Layer II with an explicit codecs parameter so it resolves
|
|
* to this decoder instead of Helix's MP3 decoder. Same pipeline as the
|
|
* embedded version, with OutputTinyGPU/I2SStream swapped for their desktop
|
|
* equivalents.
|
|
*
|
|
* File feeding uses CodecCopy, like decode-avi.ino/decode-mp4-file.ino -
|
|
* DemuxerMPG::write() may only accept part of a buffer per call, and
|
|
* (unlike StreamCopy) CodecCopy retries the remainder instead of dropping
|
|
* it, which would desync the demuxer from the real pack/PES byte stream.
|
|
*
|
|
* Audio/video sync: DemuxerMPG dispatches audio/video as fast as bytes can
|
|
* be parsed - all real pacing happens in videoSync (PacedVideoOutput,
|
|
* see Video/PacedVideoOutput.h), which sits between the demuxer and
|
|
* mpgDecoder. It buffers each decoded picture and renders it (decode +
|
|
* cv::imshow) from its own background task, timed against audioClock - an
|
|
* AudioTimeSourceStream inserted between multiDecoder and PortAudioStream
|
|
* that turns "how many decoded PCM bytes have reached the audio device so
|
|
* far" into an elapsed-ms clock, so video is paced to how far audio has
|
|
* actually played rather than a separate wall-clock schedule. OutputOpenCV
|
|
* pulls its picture size from DemuxerMPG via setVideoInfoSource(), same as
|
|
* decode-mp4-file.ino/decode-mp4.ino.
|
|
*
|
|
* Pipeline: File -> CodecCopy -> DemuxerMPG (demux)
|
|
* -> EncodedAudioStream (DecoderHelix) -> AudioTimeSourceStream (audio
|
|
* clock) -> PortAudioStream (audio)
|
|
* \-> PacedVideoOutput (buffer + schedule) -> MPGDecoder (MPEG-1
|
|
* decode -> RGB565) -> OutputOpenCV (draw) (video, own background task)
|
|
*
|
|
* To build & run:
|
|
* - mkdir build && cd build && cmake .. && make
|
|
* - ./decode-mpg
|
|
*
|
|
* @author Phil Schatzmann
|
|
* @copyright GPLv3
|
|
*/
|
|
#include "AudioTools.h"
|
|
#include "AudioTools/AudioCodecs/ContainerMPG.h"
|
|
#include "AudioTools/AudioCodecs/CodecHelix.h"
|
|
#include "AudioTools/AudioCodecs/CodecMP2.h"
|
|
#include "AudioTools/AudioLibs/PortAudioStream.h"
|
|
#include "AudioTools/Video/CodecMPG.h"
|
|
#include "AudioTools/Video/OutputOpenCV.h"
|
|
#include "SD.h"
|
|
|
|
// ---- File to play ----
|
|
const char *file_path = "/media/pschatzmann/External/Videos/output176x144-v2.mpg";
|
|
|
|
MPGDecoder mpgDecoder;
|
|
OutputOpenCV videoOut;
|
|
// 3rd arg: scheduling delay compensating for PortAudioStream's own output
|
|
// buffering - see PacedVideoOutput::setSchedulingDelayMs(); tune to
|
|
// match your actual audio config (this uses PortAudioStream's defaults).
|
|
PacedVideoOutput videoSync(mpgDecoder, 0, 50);
|
|
PortAudioStream out;
|
|
AudioTimeSourceStream audioClock(out); // decoded-PCM-bytes-based playback clock
|
|
MP2Decoder mp2Decoder;
|
|
DecoderHelix multiDecoder;
|
|
EncodedAudioStream audioOut(&audioClock, &multiDecoder); // decodes MP3/AAC/WAV -> audioClock -> PortAudio
|
|
|
|
DemuxerMPG mpgDemuxer;
|
|
File file;
|
|
CodecCopy copier(mpgDemuxer, file);
|
|
|
|
void setup() {
|
|
AudioToolsLogger.begin(Serial, AudioToolsLogLevel::Info);
|
|
|
|
// add support for mp2 audio (dedicated TinyMP2 decoder, registered
|
|
// under the specific codecs-parameterized mime DemuxerMPG reports for
|
|
// Layer II - see DemuxerMPG::mime())
|
|
multiDecoder.addDecoder(mp2Decoder, "audio/mpeg; codecs=\"mpeg1-layer2\"");
|
|
|
|
file = SD.open(file_path);
|
|
if (!file) {
|
|
Serial.print("Could not open ");
|
|
Serial.println(file_path);
|
|
exit(1);
|
|
}
|
|
|
|
auto cfg = out.defaultConfig(TX_MODE);
|
|
out.begin(cfg);
|
|
|
|
/// Define audio mime
|
|
multiDecoder.setMimeSource(mpgDemuxer);
|
|
audioOut.begin();
|
|
|
|
// mpgDecoder.setIgnorePFrames(true); // skip P-pictures, only decode I/B for speed
|
|
mpgDecoder.setOutput(videoOut);
|
|
mpgDecoder.begin();
|
|
|
|
// OutputOpenCV needs a picture size before it can display a raw
|
|
// (non-MJPEG) frame - pulls width/height from DemuxerMPG automatically
|
|
// on every write() instead of polling for it in loop().
|
|
videoOut.setVideoFormat(VideoFormat::RGB565);
|
|
videoOut.setVideoInfoSource(mpgDemuxer);
|
|
|
|
videoSync.setAudioClock(audioClock);
|
|
|
|
mpgDemuxer.setOutputAudio(audioOut);
|
|
mpgDemuxer.setOutputVideo(videoSync);
|
|
mpgDemuxer.begin();
|
|
}
|
|
|
|
void loop() {
|
|
// fps is only known once the video ES's sequence_header has streamed in
|
|
// (see DemuxerMPG::getVideoInfo()) - cheap to just keep checking until
|
|
// it's non-zero, same as the picture-size polling OutputOpenCV already
|
|
// relies on via setVideoInfoSource().
|
|
float fps = mpgDemuxer.getVideoInfo().fps;
|
|
if (fps > 0) videoSync.setFps(fps);
|
|
|
|
if (file && copier.copy()) {
|
|
// no-op - OutputOpenCV pulls its size from mpgDemuxer via
|
|
// setVideoInfoSource(), no manual polling needed here anymore
|
|
} else {
|
|
Serial.println("Done");
|
|
file.close();
|
|
exit(0);
|
|
}
|
|
}
|