Question

I have a video feed that sends me audio using the ADPCM codec. However, android only supports PCM format. How can I convert the ADPCM audio feed into a PCM audio feed?

The answer to this may be similar to the answer to this question.

I have successfully decoded the frame with this code:

int len = avcodec_decode_audio4(pAudioCodecCtx, pAudioFrame, &frameFinished, &packet);

Is the secret here to use a reverse encode function?

Here is what I have so far in my audio decode function:

<!-- language: c -->
if(packet_queue_get(env, javaThread, pAudioPacketQueue, &packet, 1) < 0) {
    LOGE("audio - after get packet failed");
    return;
}
LOGD("Dequeued audio packet");

// calculate frame size
int frameSize;
if (pPcmAudioCodecCtx->frame_size) {
    frameSize = pPcmAudioCodecCtx->frame_size;
} else {
    /* if frame_size is not set, the number of samples must be
     * calculated from the buffer size */
    int64_t nb_samples = (int64_t)AUDIO_PCM_OUTBUFF_SIZE * 8 /
            (av_get_bits_per_sample(pPcmAudioCodecCtx->codec_id) *
                    pPcmAudioCodecCtx->channels);
    frameSize = nb_samples;
}

int pcmBytesPerSample = av_get_bytes_per_sample(pPcmAudioCodecCtx->sample_fmt);
int pcmFrameBytes = frameSize * pcmBytesPerSample * pPcmAudioCodecCtx->channels;

uint8_t *pDataStart = packet.data;
while(packet.size > 0) {
    int len = avcodec_decode_audio4(pAudioCodecCtx, pAudioFrame, &frameFinished, &packet);
    LOGD("Decoded ADPCM frame");

    if (len < 0) {
        LOGE("Error while decoding audio");
        return;
    }

    if (frameFinished) {
        // store frame data in FIFO buffer
        uint8_t *inputBuffer = pAudioFrame->data[0];
        int inputBufferSize = pAudioFrame->linesize[0];
        av_fifo_generic_write(fifoBuffer, inputBuffer, inputBufferSize, NULL);
        LOGD("Added ADPCM frame to FIFO buffer");

        // check if fifo buffer has enough data for a PCM frame
        while (av_fifo_size(fifoBuffer) >= pcmFrameBytes) {
            LOGI("PCM frame data in FIFO buffer");

            // read frame's worth of data from FIFO buffer
            av_fifo_generic_read(fifoBuffer, pAudioPcmOutBuffer, pcmFrameBytes, NULL);
            LOGD("Read data from FIFO buffer into pcm frame");


            avcodec_get_frame_defaults(pPcmAudioFrame);
            LOGD("Got frame defaults");

            pPcmAudioFrame->nb_samples = pcmFrameBytes / (pPcmAudioCodecCtx->channels *
                    pcmBytesPerSample);

            avcodec_fill_audio_frame(pPcmAudioFrame, pPcmAudioCodecCtx->channels,
                    pPcmAudioCodecCtx->sample_fmt,
                    pAudioPcmOutBuffer, pcmFrameBytes, 1);
            LOGD("Filled frame audio with data");

            // fill audio play buffer
            int dataSize = pPcmAudioFrame->linesize[0];
            LOGD("Data to output: %d", dataSize);
            jbyteArray audioPlayBuffer = (jbyteArray) env->GetObjectField(ffmpegCtx, env->GetFieldID(cls, "audioPlayBuffer", "[B"));
            jbyte *bytes = env->GetByteArrayElements(audioPlayBuffer, NULL);
            memcpy(bytes, pPcmAudioFrame->data[0], dataSize);
            env->ReleaseByteArrayElements(audioPlayBuffer, bytes, 0);
            LOGD("Copied data into Java array");

            env->CallVoidMethod(player, env->GetMethodID(playerCls, "updateAudio", "(I)V"), dataSize);
        }

Answer 1

It turns out that the audio_decode_ functions return 16 bit PCM format, and that I just didn’t know how to access it properly.

Here is the altered code inside the packet loop that plays the audio based on avcodec_decode_audio4.

int len = avcodec_decode_audio4(pAudioCodecCtx, pAudioFrame, &frameFinished, &packet);

if (len < 0) {
    LOGE("Error while decoding audio");
    return;
}

if (frameFinished) {
    int planeSize;
    uint8_t *pcmBuffer = pAudioFrame->extended_data[0];
    int dataSize = av_samples_get_buffer_size(&planeSize, pAudioCodecCtx->channels,
                                                       pAudioFrame->nb_samples,
                                                       pAudioCodecCtx->sample_fmt, 1);
    // fill audio play buffer
    jbyteArray audioPlayBuffer = (jbyteArray) env->GetObjectField(ffmpegCtx, env->GetFieldID(cls, "audioPlayBuffer", "[B"));
    jbyte *bytes = env->GetByteArrayElements(audioPlayBuffer, NULL);
    memcpy(bytes, pcmBuffer, dataSize);
    env->ReleaseByteArrayElements(audioPlayBuffer, bytes, 0);
    env->CallVoidMethod(player, env->GetMethodID(playerCls, "updateAudio", "(I)V"), dataSize);
}