Android-將MediaMuxer與MediaExtractor和PCM流一起使用會導致視頻幀損壞

Question

我正在為支持流式傳輸和錄制視頻的應用程序編寫類。 簡而言之，當電話流式傳輸和錄制時，使用MediaRecorder將音頻保存在PCM文件中，將視頻保存在mp4文件中。 我的目標是在錄制完成后使用MediaMuxer並將兩個輸入合並到一個新的合並的.mp4文件中。

我嘗試使用MediaMuxer編碼音頻並使用MediaExtractor提取視頻。 原始視頻和音頻文件均完整無缺，並且輸出文件包含正確的音頻，但視頻似乎已損壞，就像跳過了幀一樣。

這是我當前正在使用的代碼：

public class StreamRecordingMuxer {

    private static final String TAG = StreamRecordingMuxer.class.getSimpleName();

    private static final String COMPRESSED_AUDIO_FILE_MIME_TYPE = "audio/mp4a-latm";

    private static final int CODEC_TIMEOUT = 5000;

    private int bitrate;
    private int sampleRate;
    private int channelCount;

    // Audio state
    private MediaFormat audioFormat;
    private MediaCodec mediaCodec;
    private MediaMuxer mediaMuxer;
    private ByteBuffer[] codecInputBuffers;
    private ByteBuffer[] codecOutputBuffers;
    private MediaCodec.BufferInfo audioBufferInfo;
    private String outputPath;
    private int audioTrackId;
    private int totalBytesRead;
    private double presentationTimeUs;

    // Video state
    private int videoTrackId;
    private MediaExtractor videoExtractor;
    private MediaFormat videoFormat;
    private String videoPath;
    private int videoTrackIndex;
    private int frameMaxInputSize;
    private int rotationDegrees;

    public StreamRecordingMuxer(final int bitrate, final int sampleRate, int channelCount) {
        this.bitrate = bitrate;
        this.sampleRate = sampleRate;
        this.channelCount = channelCount;
    }

    public void setOutputPath(final String outputPath) {
        this.outputPath = outputPath;
    }

    public void setVideoPath(String videoPath) {
        this.videoPath = videoPath;
    }

    public void prepare() {
        if (outputPath == null) {
            throw new IllegalStateException("The output path must be set first!");
        }
        try {
            audioFormat = MediaFormat.createAudioFormat(COMPRESSED_AUDIO_FILE_MIME_TYPE, sampleRate, channelCount);
            audioFormat.setInteger(MediaFormat.KEY_AAC_PROFILE, MediaCodecInfo.CodecProfileLevel.AACObjectLC);
            audioFormat.setInteger(MediaFormat.KEY_BIT_RATE, bitrate);

            if (videoPath != null) {
                videoExtractor = new MediaExtractor();
                videoExtractor.setDataSource(videoPath);
                videoFormat = findVideoFormat(videoExtractor);
            }

            mediaCodec = MediaCodec.createEncoderByType(COMPRESSED_AUDIO_FILE_MIME_TYPE);
            mediaCodec.configure(audioFormat, null, null, MediaCodec.CONFIGURE_FLAG_ENCODE);
            mediaCodec.start();

            codecInputBuffers = mediaCodec.getInputBuffers();
            codecOutputBuffers = mediaCodec.getOutputBuffers();

            audioBufferInfo = new MediaCodec.BufferInfo();

            mediaMuxer = new MediaMuxer(outputPath, MediaMuxer.OutputFormat.MUXER_OUTPUT_MPEG_4);

            if (videoPath != null) {
                videoTrackId = mediaMuxer.addTrack(videoFormat);
                mediaMuxer.setOrientationHint(rotationDegrees);
            }

            totalBytesRead = 0;
            presentationTimeUs = 0;
        } catch (IOException e) {
            Log.e(TAG, "Exception while initializing StreamRecordingMuxer", e);
        }
    }

    public void stop() {
        Log.d(TAG, "Stopping StreamRecordingMuxer");
        handleEndOfStream();

        mediaCodec.stop();
        mediaCodec.release();
        mediaMuxer.stop();
        mediaMuxer.release();

        if (videoExtractor != null) {
            videoExtractor.release();
        }
    }

    private void handleEndOfStream() {
        int inputBufferIndex = mediaCodec.dequeueInputBuffer(CODEC_TIMEOUT);
        mediaCodec.queueInputBuffer(inputBufferIndex, 0, 0, (long) presentationTimeUs, MediaCodec.BUFFER_FLAG_END_OF_STREAM);
        writeAudioOutputs();
    }

    private MediaFormat findVideoFormat(MediaExtractor extractor) {
        MediaFormat videoFormat;
        int videoTrackCount = extractor.getTrackCount();
        for (int i = 0; i < videoTrackCount; i++) {
            videoFormat = extractor.getTrackFormat(i);
            Log.d(TAG, "Video Format " + videoFormat.toString());
            String mimeType = videoFormat.getString(MediaFormat.KEY_MIME);
            if (mimeType.startsWith("video/")) {
                videoTrackIndex = i;
                frameMaxInputSize = videoFormat.getInteger(MediaFormat.KEY_MAX_INPUT_SIZE);
                rotationDegrees = videoFormat.getInteger(MediaFormat.KEY_ROTATION);
                // frameRate = videoFormat.getInteger(MediaFormat.KEY_FRAME_RATE);
                // videoDuration = videoFormat.getLong(MediaFormat.KEY_DURATION);
                return videoFormat;
            }
        }

        return null;
    }

    private void writeVideoToMuxer() {
        ByteBuffer buffer = ByteBuffer.allocate(frameMaxInputSize);
        MediaCodec.BufferInfo videoBufferInfo = new MediaCodec.BufferInfo();
        videoExtractor.unselectTrack(videoTrackIndex);
        videoExtractor.selectTrack(videoTrackIndex);

        while (true) {
            buffer.clear();
            int sampleSize = videoExtractor.readSampleData(buffer, 0);
            if (sampleSize < 0) {
                videoExtractor.unselectTrack(videoTrackIndex);
                break;
            }

            videoBufferInfo.size = sampleSize;
            videoBufferInfo.presentationTimeUs = videoExtractor.getSampleTime();
            videoBufferInfo.flags = videoExtractor.getSampleFlags();

            mediaMuxer.writeSampleData(videoTrackId, buffer, videoBufferInfo);
            videoExtractor.advance();
        }
    }

    private void encodeAudioPCM(InputStream is) throws IOException {
        byte[] tempBuffer = new byte[2 * sampleRate];
        boolean hasMoreData = true;
        boolean stop = false;

        while (!stop) {
            int inputBufferIndex = 0;
            int currentBatchRead = 0;
            while (inputBufferIndex != -1 && hasMoreData && currentBatchRead <= 50 * sampleRate) {
                inputBufferIndex = mediaCodec.dequeueInputBuffer(CODEC_TIMEOUT);

                if (inputBufferIndex >= 0) {
                    ByteBuffer buffer = codecInputBuffers[inputBufferIndex];
                    buffer.clear();

                    int bytesRead = is.read(tempBuffer, 0, buffer.limit());
                    if (bytesRead == -1) {
                        mediaCodec.queueInputBuffer(inputBufferIndex, 0, 0, (long) presentationTimeUs, 0);
                        hasMoreData = false;
                        stop = true;
                    } else {
                        totalBytesRead += bytesRead;
                        currentBatchRead += bytesRead;
                        buffer.put(tempBuffer, 0, bytesRead);
                        mediaCodec.queueInputBuffer(inputBufferIndex, 0, bytesRead, (long) presentationTimeUs, 0);
                        presentationTimeUs = 1000000L * (totalBytesRead / 2) / sampleRate;
                    }
                }
            }

            writeAudioOutputs();
        }

        is.close();
    }

    public void start(InputStream inputStream) throws IOException {
        Log.d(TAG, "Starting encoding of InputStream");
        encodeAudioPCM(inputStream);
        Log.d(TAG, "Finished encoding of InputStream");

        if (videoPath != null) {
            writeVideoToMuxer();
        }
    }

    private void writeAudioOutputs() {
        int outputBufferIndex = 0;
        while (outputBufferIndex != MediaCodec.INFO_TRY_AGAIN_LATER) {
            outputBufferIndex = mediaCodec.dequeueOutputBuffer(audioBufferInfo, CODEC_TIMEOUT);
            if (outputBufferIndex >= 0) {
                ByteBuffer encodedData = codecOutputBuffers[outputBufferIndex];
                encodedData.position(audioBufferInfo.offset);
                encodedData.limit(audioBufferInfo.offset + audioBufferInfo.size);

                if ((audioBufferInfo.flags & MediaCodec.BUFFER_FLAG_CODEC_CONFIG) != 0 && audioBufferInfo.size != 0) {
                    mediaCodec.releaseOutputBuffer(outputBufferIndex, false);
                } else {
                    mediaMuxer.writeSampleData(audioTrackId, codecOutputBuffers[outputBufferIndex], audioBufferInfo);
                    mediaCodec.releaseOutputBuffer(outputBufferIndex, false);
                }
            } else if (outputBufferIndex == MediaCodec.INFO_OUTPUT_FORMAT_CHANGED) {
                audioFormat = mediaCodec.getOutputFormat();
                audioTrackId = mediaMuxer.addTrack(audioFormat);
                mediaMuxer.start();
            }
        }
    }
}

Answer 1

我終於設法找到了與實際Muxer代碼無關的答案：事實證明，在創建音頻文件時，演示時間計算錯誤。