ZLMediaKit/ext-codec/AAC.cpp

/*
 * Copyright (c) 2016-present The ZLMediaKit project authors. All Rights Reserved.
 *
 * This file is part of ZLMediaKit(https://github.com/ZLMediaKit/ZLMediaKit).
 *
 * Use of this source code is governed by MIT-like license that can be found in the
 * LICENSE file in the root of the source tree. All contributing project authors
 * may be found in the AUTHORS file in the root of the source tree.
 */

#include "AAC.h"
#include "AACRtp.h"
#include "AACRtmp.h"
#include "Common/Parser.h"
#include "Extension/Factory.h"
#ifdef ENABLE_MP4
#include "mpeg4-aac.h"
#endif

using namespace std;
using namespace toolkit;

namespace mediakit{

#ifndef ENABLE_MP4
unsigned const samplingFrequencyTable[16] = { 96000, 88200, 64000, 48000, 44100, 32000, 24000, 22050, 16000, 12000, 11025, 8000, 7350, 0, 0, 0 };

class AdtsHeader {
public:
    unsigned int syncword = 0; // 12 bslbf 同步字The bit string ‘1111 1111 1111’，说明一个ADTS帧的开始
    unsigned int id; // 1 bslbf   MPEG 标示符, 设置为1
    unsigned int layer; // 2 uimsbf Indicates which layer is used. Set to ‘00’
    unsigned int protection_absent; // 1 bslbf  表示是否误码校验
    unsigned int profile; // 2 uimsbf  表示使用哪个级别的AAC，如01 Low Complexity(LC)--- AACLC
    unsigned int sf_index; // 4 uimsbf  表示使用的采样率下标
    unsigned int private_bit; // 1 bslbf
    unsigned int channel_configuration; // 3 uimsbf  表示声道数
    unsigned int original; // 1 bslbf
    unsigned int home; // 1 bslbf
    // 下面的为改变的参数即每一帧都不同
    unsigned int copyright_identification_bit; // 1 bslbf
    unsigned int copyright_identification_start; // 1 bslbf
    unsigned int aac_frame_length; // 13 bslbf  一个ADTS帧的长度包括ADTS头和raw data block
    unsigned int adts_buffer_fullness; // 11 bslbf     0x7FF 说明是码率可变的码流
    // no_raw_data_blocks_in_frame 表示ADTS帧中有number_of_raw_data_blocks_in_frame + 1个AAC原始帧.
    // 所以说number_of_raw_data_blocks_in_frame == 0
    // 表示说ADTS帧中有一个AAC数据块并不是说没有。(一个AAC原始帧包含一段时间内1024个采样及相关数据)
    unsigned int no_raw_data_blocks_in_frame; // 2 uimsfb
};

static void dumpAdtsHeader(const AdtsHeader &hed, uint8_t *out) {
    out[0] = (hed.syncword >> 4 & 0xFF); // 8bit
    out[1] = (hed.syncword << 4 & 0xF0); // 4 bit
    out[1] |= (hed.id << 3 & 0x08); // 1 bit
    out[1] |= (hed.layer << 1 & 0x06); // 2bit
    out[1] |= (hed.protection_absent & 0x01); // 1 bit
    out[2] = (hed.profile << 6 & 0xC0); // 2 bit
    out[2] |= (hed.sf_index << 2 & 0x3C); // 4bit
    out[2] |= (hed.private_bit << 1 & 0x02); // 1 bit
    out[2] |= (hed.channel_configuration >> 2 & 0x03); // 1 bit
    out[3] = (hed.channel_configuration << 6 & 0xC0); // 2 bit
    out[3] |= (hed.original << 5 & 0x20); // 1 bit
    out[3] |= (hed.home << 4 & 0x10); // 1 bit
    out[3] |= (hed.copyright_identification_bit << 3 & 0x08); // 1 bit
    out[3] |= (hed.copyright_identification_start << 2 & 0x04); // 1 bit
    out[3] |= (hed.aac_frame_length >> 11 & 0x03); // 2 bit
    out[4] = (hed.aac_frame_length >> 3 & 0xFF); // 8 bit
    out[5] = (hed.aac_frame_length << 5 & 0xE0); // 3 bit
    out[5] |= (hed.adts_buffer_fullness >> 6 & 0x1F); // 5 bit
    out[6] = (hed.adts_buffer_fullness << 2 & 0xFC); // 6 bit
    out[6] |= (hed.no_raw_data_blocks_in_frame & 0x03); // 2 bit
}

static bool parseAacConfig(const string &config, AdtsHeader &adts) {
    if (config.size() < 2) {
        return false;
    }
    uint8_t cfg1 = config[0];
    uint8_t cfg2 = config[1];

    int audioObjectType;
    int sampling_frequency_index;
    int channel_configuration;

    audioObjectType = cfg1 >> 3;
    sampling_frequency_index = ((cfg1 & 0x07) << 1) | (cfg2 >> 7);
    channel_configuration = (cfg2 & 0x7F) >> 3;

    adts.syncword = 0x0FFF;
    adts.id = 0;
    adts.layer = 0;
    adts.protection_absent = 1;
    adts.profile = audioObjectType - 1;
    adts.sf_index = sampling_frequency_index;
    adts.private_bit = 0;
    adts.channel_configuration = channel_configuration;
    adts.original = 0;
    adts.home = 0;
    adts.copyright_identification_bit = 0;
    adts.copyright_identification_start = 0;
    adts.aac_frame_length = 7;
    adts.adts_buffer_fullness = 2047;
    adts.no_raw_data_blocks_in_frame = 0;
    return true;
}
#endif// ENABLE_MP4

int getAacFrameLength(const uint8_t *data, size_t bytes) {
    uint16_t len;
    if (bytes < 7) return -1;
    if (0xFF != data[0] || 0xF0 != (data[1] & 0xF0)) {
        return -1;
    }
    len = ((uint16_t) (data[3] & 0x03) << 11) | ((uint16_t) data[4] << 3) | ((uint16_t) (data[5] >> 5) & 0x07);
    return len;
}

string makeAacConfig(const uint8_t *hex, size_t length){
#ifndef ENABLE_MP4
    if (!(hex[0] == 0xFF && (hex[1] & 0xF0) == 0xF0)) {
        return "";
    }
    // Get and check the 'profile':
    unsigned char profile = (hex[2] & 0xC0) >> 6; // 2 bits
    if (profile == 3) {
        return "";
    }

    // Get and check the 'sampling_frequency_index':
    unsigned char sampling_frequency_index = (hex[2] & 0x3C) >> 2; // 4 bits
    if (samplingFrequencyTable[sampling_frequency_index] == 0) {
        return "";
    }

    // Get and check the 'channel_configuration':
    unsigned char channel_configuration = ((hex[2] & 0x01) << 2) | ((hex[3] & 0xC0) >> 6); // 3 bits
    unsigned char audioSpecificConfig[2];
    unsigned char const audioObjectType = profile + 1;
    audioSpecificConfig[0] = (audioObjectType << 3) | (sampling_frequency_index >> 1);
    audioSpecificConfig[1] = (sampling_frequency_index << 7) | (channel_configuration << 3);
    return string((char *)audioSpecificConfig,2);
#else
    struct mpeg4_aac_t aac;
    memset(&aac, 0, sizeof(aac));
    if (mpeg4_aac_adts_load(hex, length, &aac) > 0) {
        char buf[32] = {0};
        int len = mpeg4_aac_audio_specific_config_save(&aac, (uint8_t *) buf, sizeof(buf));
        if (len > 0) {
            return string(buf, len);
        }
    }
    WarnL << "生成aac config失败, adts header:" << hexdump(hex, length);
    return "";
#endif
}

int dumpAacConfig(const string &config, size_t length, uint8_t *out, size_t out_size) {
#ifndef ENABLE_MP4
    AdtsHeader header;
    parseAacConfig(config, header);
    header.aac_frame_length = (decltype(header.aac_frame_length))(ADTS_HEADER_LEN + length);
    dumpAdtsHeader(header, out);
    return ADTS_HEADER_LEN;
#else
    struct mpeg4_aac_t aac;
    memset(&aac, 0, sizeof(aac));
    int ret = mpeg4_aac_audio_specific_config_load((uint8_t *) config.data(), config.size(), &aac);
    if (ret > 0) {
        ret = mpeg4_aac_adts_save(&aac, length, out, out_size);
    }
    if (ret < 0) {
        WarnL << "生成adts头失败:" << ret << ", aac config:" << hexdump(config.data(), config.size());
    }
    assert((int)out_size >= ret);
    return ret;
#endif
}

bool parseAacConfig(const string &config, int &samplerate, int &channels) {
#ifndef ENABLE_MP4
    AdtsHeader header;
    if (!parseAacConfig(config, header)) {
        return false;
    }
    samplerate = samplingFrequencyTable[header.sf_index];
    channels = header.channel_configuration;
    return true;
#else
    struct mpeg4_aac_t aac;
    memset(&aac, 0, sizeof(aac));
    int ret = mpeg4_aac_audio_specific_config_load((uint8_t *) config.data(), config.size(), &aac);
    if (ret > 0) {
        samplerate = aac.sampling_frequency;
        channels = aac.channels;
        return true;
    }
    WarnL << "获取aac采样率、声道数失败:" << hexdump(config.data(), config.size());
    return false;
#endif
}

////////////////////////////////////////////////////////////////////////////////////////////////////

/**
 * aac类型SDP
 */
class AACSdp : public Sdp {
public:
    /**
     * 构造函数
     * @param aac_cfg aac两个字节的配置描述
     * @param payload_type rtp payload type
     * @param sample_rate 音频采样率
     * @param channels 通道数
     * @param bitrate 比特率
     */
    AACSdp(const string &aac_cfg, int payload_type, int sample_rate, int channels, int bitrate)
        : Sdp(sample_rate, payload_type) {
        _printer << "m=audio 0 RTP/AVP " << payload_type << "\r\n";
        if (bitrate) {
            _printer << "b=AS:" << bitrate << "\r\n";
        }
        _printer << "a=rtpmap:" << payload_type << " " << getCodecName(CodecAAC) << "/" << sample_rate << "/" << channels << "\r\n";

        string configStr;
        char buf[4] = { 0 };
        for (auto &ch : aac_cfg) {
            snprintf(buf, sizeof(buf), "%02X", (uint8_t)ch);
            configStr.append(buf);
        }
        _printer << "a=fmtp:" << payload_type << " streamtype=5;profile-level-id=1;mode=AAC-hbr;"
                 << "sizelength=13;indexlength=3;indexdeltalength=3;config=" << configStr << "\r\n";
    }

    string getSdp() const override { return _printer; }

private:
    _StrPrinter _printer;
};

////////////////////////////////////////////////////////////////////////////////////////////////////

AACTrack::AACTrack(const string &aac_cfg) {
    if (aac_cfg.size() < 2) {
        throw std::invalid_argument("adts配置必须最少2个字节");
    }
    _cfg = aac_cfg;
    update();
}

CodecId AACTrack::getCodecId() const {
    return CodecAAC;
}

bool AACTrack::ready() const {
    return !_cfg.empty();
}

int AACTrack::getAudioSampleRate() const {
    return _sampleRate;
}

int AACTrack::getAudioSampleBit() const {
    return _sampleBit;
}

int AACTrack::getAudioChannel() const {
    return _channel;
}

static Frame::Ptr addADTSHeader(const Frame::Ptr &frame_in, const std::string &aac_config) {
    auto frame = FrameImp::create();
    frame->_codec_id = CodecAAC;
    // 生成adts头
    char adts_header[32] = { 0 };
    auto size = dumpAacConfig(aac_config, frame_in->size(), (uint8_t *)adts_header, sizeof(adts_header));
    CHECK(size > 0, "Invalid adts config");
    frame->_prefix_size = size;
    frame->_dts = frame_in->dts();
    frame->_buffer.assign(adts_header, size);
    frame->_buffer.append(frame_in->data(), frame_in->size());
    frame->setIndex(frame_in->getIndex());
    return frame;
}

bool AACTrack::inputFrame(const Frame::Ptr &frame) {
    if (!frame->prefixSize()) {
        CHECK(ready());
        return inputFrame_l(addADTSHeader(frame, _cfg));
    }

    bool ret = false;
    //有adts头，尝试分帧
    int64_t dts = frame->dts();
    int64_t pts = frame->pts();

    auto ptr = frame->data();
    auto end = frame->data() + frame->size();
    while (ptr < end) {
        auto frame_len = getAacFrameLength((uint8_t *)ptr, end - ptr);
        if (frame_len < ADTS_HEADER_LEN) {
            break;
        }
        if (frame_len == (int)frame->size()) {
            return inputFrame_l(frame);
        }
        auto sub_frame = std::make_shared<FrameInternalBase<FrameFromPtr>>(frame, (char *)ptr, frame_len, dts, pts, ADTS_HEADER_LEN);
        ptr += frame_len;
        if (ptr > end) {
            WarnL << "invalid aac length in adts header: " << frame_len
                  << ", remain data size: " << end - (ptr - frame_len);
            break;
        }
        if (inputFrame_l(sub_frame)) {
            ret = true;
        }
        dts += 1024 * 1000 / getAudioSampleRate();
        pts += 1024 * 1000 / getAudioSampleRate();
    }
    return ret;
}

bool AACTrack::inputFrame_l(const Frame::Ptr &frame) {
    if (_cfg.empty() && frame->prefixSize()) {
        // 未获取到aac_cfg信息，根据7个字节的adts头生成aac config
        _cfg = makeAacConfig((uint8_t *)(frame->data()), frame->prefixSize());
        update();
    }

    if (frame->size() > frame->prefixSize()) {
        // 除adts头外，有实际负载
        return AudioTrack::inputFrame(frame);
    }
    return false;
}

toolkit::Buffer::Ptr AACTrack::getExtraData() const {
    CHECK(ready());
    return std::make_shared<BufferString>(_cfg);
}

void AACTrack::setExtraData(const uint8_t *data, size_t size) {
    CHECK(size >= 2);
    _cfg.assign((char *)data, size);
    update();
}

bool AACTrack::update() {
    return parseAacConfig(_cfg, _sampleRate, _channel);
}

Track::Ptr AACTrack::clone() const {
    return std::make_shared<AACTrack>(*this);
}

Sdp::Ptr AACTrack::getSdp(uint8_t payload_type) const {
    if (!ready()) {
        WarnL << getCodecName() << " Track未准备好";
        return nullptr;
    }
    return std::make_shared<AACSdp>(getExtraData()->toString(), payload_type, getAudioSampleRate(), getAudioChannel(), getBitRate() / 1024);
}

namespace {

CodecId getCodec() {
    return CodecAAC;
}

Track::Ptr getTrackByCodecId(int sample_rate, int channels, int sample_bit) {
    return std::make_shared<AACTrack>();
}

Track::Ptr getTrackBySdp(const SdpTrack::Ptr &track) {
    string aac_cfg_str = findSubString(track->_fmtp.data(), "config=", ";");
    if (aac_cfg_str.empty()) {
        aac_cfg_str = findSubString(track->_fmtp.data(), "config=", nullptr);
    }
    if (aac_cfg_str.empty()) {
        // 如果sdp中获取不到aac config信息，那么在rtp也无法获取，那么忽略该Track
        return nullptr;
    }
    string aac_cfg;
    for (size_t i = 0; i < aac_cfg_str.size() / 2; ++i) {
        unsigned int cfg;
        sscanf(aac_cfg_str.substr(i * 2, 2).data(), "%02X", &cfg);
        cfg &= 0x00FF;
        aac_cfg.push_back((char)cfg);
    }
    return std::make_shared<AACTrack>(aac_cfg);
}

RtpCodec::Ptr getRtpEncoderByCodecId(uint8_t pt) {
    return std::make_shared<AACRtpEncoder>();
}

RtpCodec::Ptr getRtpDecoderByCodecId() {
    return std::make_shared<AACRtpDecoder>();
}

RtmpCodec::Ptr getRtmpEncoderByTrack(const Track::Ptr &track) {
    return std::make_shared<AACRtmpEncoder>(track);
}

RtmpCodec::Ptr getRtmpDecoderByTrack(const Track::Ptr &track) {
    return std::make_shared<AACRtmpDecoder>(track);
}

size_t aacPrefixSize(const char *data, size_t bytes) {
    uint8_t *ptr = (uint8_t *)data;
    size_t prefix = 0;
    if (!(bytes > ADTS_HEADER_LEN && ptr[0] == 0xFF && (ptr[1] & 0xF0) == 0xF0)) {
        return 0;
    }
    return ADTS_HEADER_LEN;
}

Frame::Ptr getFrameFromPtr(const char *data, size_t bytes, uint64_t dts, uint64_t pts) {
    return std::make_shared<FrameFromPtr>(CodecAAC, (char *)data, bytes, dts, pts, aacPrefixSize(data, bytes));
}

} // namespace

CodecPlugin aac_plugin = { getCodec,
                           getTrackByCodecId,
                           getTrackBySdp,
                           getRtpEncoderByCodecId,
                           getRtpDecoderByCodecId,
                           getRtmpEncoderByTrack,
                           getRtmpDecoderByTrack,
                           getFrameFromPtr };

} // namespace mediakit