這類的文章太多太多,也沒時(shí)間整理,直接上部分核心源碼和注意事項(xiàng)

1：MediaCodec核心類, 在往MediaCodec中不斷的推數(shù)據(jù)時(shí)一定要使用BytePool字節(jié)數(shù)組池,MediaCodec編碼后的byte[]可以循環(huán)重復(fù)使用,避免造成內(nèi)存的抖動(dòng)

/**
 * Created by you on 2018-05-10.
 * MediaCodec核心編碼器
 */
public final class MediaEncoder implements Runnable {
    //...偽代
    //字節(jié)池
    private final BytePool bytePool;

    @Override
    public void run() {
        bufferInfo = new MediaCodec.BufferInfo();
        callback.onInitStart();
        while (isCoding.get()) {
            try {
                byte[] buffer = bufferQueue.take();
                if (buffer == null || buffer.length == 0) {
                    break;//用空byte[]來終止循環(huán)與阻塞
                }
                codecDatas(buffer);
                //緩存
                bytePool.put(buffer);
            } catch (InterruptedException e) {
                if (!isCoding.get()) {
                    break;
                }
            }
        }
        release();
    }

    /**
     * 編碼datas數(shù)據(jù)
     * @param buffer
     */
    private void codecDatas(byte[] buffer) {
        //加入緩沖區(qū), -1如果當(dāng)前沒有可用的緩沖時(shí)會(huì)進(jìn)入阻塞狀態(tài), 0時(shí)會(huì)立刻返回
        int index = mediaCodec.dequeueInputBuffer(-1);
        if (index >= 0) {
            //填充數(shù)據(jù)
            ByteBuffer inputBuffer = mediaCodec.getInputBuffer(index);
            inputBuffer.clear();
            inputBuffer.put(buffer, 0, buffer.length);

            callback.onEncodeInputBuffer(mediaCodec, buffer, index);
        }
        int encodeStatus;
        while (true) {
            //返回的三種狀態(tài) INFO_TRY_AGAIN_LATER, INFO_OUTPUT_FORMAT_CHANGED, INFO_OUTPUT_BUFFERS_CHANGED,
            encodeStatus = mediaCodec.dequeueOutputBuffer(bufferInfo, timeoutUs);
            if (encodeStatus == MediaCodec.INFO_TRY_AGAIN_LATER) {
                break;//稍后重試
            } else if (encodeStatus == MediaCodec.INFO_OUTPUT_FORMAT_CHANGED){
                //這里只會(huì)回調(diào)一次用于初始化
                callback.onFormatChanged(mediaCodec);
            } else if (encodeStatus == MediaCodec.INFO_OUTPUT_BUFFERS_CHANGED) {
                //忽略
            } else {
                //正常編碼獲得緩沖下標(biāo)
                ByteBuffer encodeData = mediaCodec.getOutputBuffer(encodeStatus);
                //寫入編碼后的數(shù)據(jù)
                callback.onWriteData(bufferInfo, encodeData);
                //釋放緩存沖,后續(xù)可以存放新的編碼后的數(shù)據(jù)
                mediaCodec.releaseOutputBuffer(encodeStatus, false);
            }
        }
    }
}

2：使用OutputStream方式寫入編碼的h264(推薦使用MediaMuxer)

ByteBuffer outputBuffer = outputBuffers[outputBufferIndex];
byte[] outData = new byte[bufferInfo.size];
outputBuffer.get(outData);
OutputStream.write(outData, 0, outData.length);

上面這種寫法雖然方便但是也有錯(cuò)誤,MediaCodec每次編碼后的幀大小可能不一樣,但是每次編碼后都new byte也還是會(huì)造成內(nèi)存抖動(dòng),需要采用如下方式,建一個(gè)適當(dāng)大小的byte緩沖,用這個(gè)緩沖去接每一幀ByteBuffere中的數(shù)據(jù)再寫入到OutputStream中, 也可以直接使用WritableByteChannel寫入，原理其實(shí)差不多

    //寫入數(shù)據(jù)緩沖
    private byte[] writeBuffer;

    @Override
    public void onWriteData(MediaCodec.BufferInfo bufferInfo, ByteBuffer encodeData) {
        if (bufferInfo.size != 0) {
            //將ByteBuffer中的數(shù)據(jù)寫到文件中
//            LogUtils.i("write buffinfosize  %d", bufferInfo.size);
            int offset = bufferInfo.offset;
            int bufferSize = bufferInfo.size;
            while (bufferSize > writeBuffer.length) {
                writeByteBuffer(encodeData, offset, writeBuffer.length);
                bufferSize -= writeBuffer.length;
                offset += writeBuffer.length;
            }
            if (bufferSize > 0) {
                writeByteBuffer(encodeData, offset, bufferSize);
            }
            //byte[] buf = new byte[bufferInfo.size];
            //encodeData.get(buf); 不能用此種方式寫入,內(nèi)存抖動(dòng)極大
        }
    }

    /**
     * 將ByteBuffer通過byte[]寫入到文件
     * @param encodeData
     * @param offset
     * @param length
     */
    private void writeByteBuffer(ByteBuffer encodeData, int offset, int length) {
        encodeData.position(offset);
        encodeData.limit(offset + length);

        encodeData.get(writeBuffer, 0, length);
        try {
            bos.write(writeBuffer, 0, length);
        } catch (IOException e) {
            e.printStackTrace();
        }
    }

3：使用OutputStream方式寫入編碼后aac文件(推薦使用MediaMuxer),同h264一樣需要一個(gè)byte緩沖去接ByteBuffere中的數(shù)據(jù)再寫入到OutputStream中
aac壓縮格式可以直接使用播放器播放,采用 ADTS 格式需要給每幀加上 7 個(gè)字節(jié)的頭信息.(MediaMuxer會(huì)自動(dòng)處理)

  @Override
    public void onWriteData(MediaCodec.BufferInfo bufferInfo, ByteBuffer encodeData) {
        if (bufferInfo.size != 0) {
            encodeData.position(bufferInfo.offset);
            encodeData.limit(bufferInfo.offset + bufferInfo.size);

            addADTStoPacket(bufferInfo.size + 7);
            try {
                bos.write(adtsHeader, 0, 7);
            } catch (IOException e) {
                e.printStackTrace();
            }

            //將ByteBuffer中的數(shù)據(jù)寫到文件中
            LogUtils.i("write buffinfosize  %d", bufferInfo.size);
            int offset = bufferInfo.offset;
            int bufferSize = bufferInfo.size;
            while (bufferSize > writeBuffer.length) {
                writeByteBuffer(encodeData, offset, writeBuffer.length);
                bufferSize -= writeBuffer.length;
                offset += writeBuffer.length;
            }
            if (bufferSize > 0) {
                writeByteBuffer(encodeData, offset, bufferSize);
            }
            //byte[] buf = new byte[bufferInfo.size];
            //encodeData.get(buf); 不能用此種方式寫入,內(nèi)存抖動(dòng)極大
        }
    }

    private void addADTStoPacket(int packetLen) {
        adtsHeader[3] = (byte) (((chanCfg & 3) << 6) + (packetLen >> 11));
        adtsHeader[4] = (byte) ((packetLen & 0x7FF) >> 3);
        adtsHeader[5] = (byte) (((packetLen & 7) << 5) + 0x1F);
        adtsHeader[6] = (byte) 0xFC;
    }

4：MediaMuxer混合錄制,需要aac與h264都有MediaCodec回調(diào)INFO_OUTPUT_FORMAT_CHANGED狀態(tài)時(shí)添加mediaMuxer.addTrack(mediaCodec.getOutputFormat());才可以開啟MediaMuxer携狭， 結(jié)束亦是如此法褥，否則會(huì)拋異常
會(huì)用到多線程之間的操作,最先mediaMuxer.addTrack的進(jìn)入wait()等侍狀態(tài)酣藻，最后一個(gè)mediaMuxer.addTrack的notifyAll()

  /**
     * avc與aac同時(shí)都已a(bǔ)ddTrack時(shí)才可開啟
     */
    private synchronized void startMuxer() {
        if (!isMuxerStarted && isRecording) {
            if (audioTrackIndex != -1 && h264TrackIndex != -1) {
                mediaMuxer.start();
                isMuxerStarted = true;
                //最后一個(gè)addTrack的來開啟
                notifyAll();
            } else {
                long c = System.currentTimeMillis();
                do {
                    try {
                        //先addTrack的處于等侍狀態(tài)
                        wait();
                    } catch (InterruptedException e) {
                        e.printStackTrace();
                    }
                } while (isRecording && (audioTrackIndex == -1 || h264TrackIndex == -1));
                long n = System.currentTimeMillis() - c;
                LogUtils.i("wait... %d", n);
            }
        }
    }

    /**
     * aac, h264編碼都停止時(shí)才可停止
     */
    private synchronized void stopMuxer() {
        if (isMuxerStarted && h264Released && audioReleased) {
            mediaMuxer.stop();
            mediaMuxer.release();
            isMuxerStarted = false;
            mediaMuxer = null;
            LogUtils.i("mp4recorder release...");
        }
    }

5：注意在MediaCodec編碼aac時(shí),需要對(duì)音頻進(jìn)行時(shí)間采樣計(jì)算,否則在混合的時(shí)候容易拋出類似
MPEG4Writer: timestampUs 6220411 < lastTimestampUs 6220442 for Audio track異常,報(bào)異常原因也很簡單,當(dāng)aac編碼通過mediaMuxer.addTrack()進(jìn)入等侍wait狀態(tài)等侍h264同步進(jìn)行時(shí),等侍的時(shí)間超過了差值6220442微秒, 另外在視頻H264與PCM解碼播放時(shí)锭沟，一般也都是根據(jù)pcm的播放速度來做同步的尔当，因此音頻的時(shí)間采樣非常非常重要

public final class AudioPresentationTime {
    private long startTime;
    private final long bufferDurationUs;
    private long currentCount;
    /**
     *
     * @param bufferSize
     * @param sampleRate
     * @param channelCount
     * @param audioFormat
     */
    public AudioPresentationTime(int bufferSize, int sampleRate, int channelCount, int audioFormat) {
        int bitByteSize = audioFormat == AudioFormat.ENCODING_PCM_16BIT ? 2 : 1; //16bit = 2 byte
        bufferDurationUs = 1_000_000L * (bufferSize / (channelCount * bitByteSize)) / sampleRate;
    }

    public void start() {
        startTime = System.nanoTime() / 1000L;
        currentCount = 0;
    }

    public long getPresentationTimeUs() {
        return currentCount++ * bufferDurationUs + startTime;
    }
}

補(bǔ)充：除了要對(duì)對(duì)音頻進(jìn)行時(shí)間采樣計(jì)算外,H264編碼的Camera采集數(shù)據(jù)也需要注意預(yù)覽的尺寸大小和幀率的控制, 在預(yù)覽尺寸過大或者幀率過高的時(shí)候,MediaCodec編碼的速度趕不上采集的速度(預(yù)覽回調(diào)的YUV數(shù)據(jù)越大編碼耗時(shí)越大),就會(huì)造成等侍編碼的YUV數(shù)據(jù)隊(duì)列的爆滿內(nèi)存溢出坞靶，或者在有限制隊(duì)列大小時(shí)也會(huì)造成中間幀的丟失.

附上源碼https://github.com/youxiaochen/CameraMedia