jazzlost

Share & Discuss

2020/02/23 · 李AA

关于Wave格式

Study Of Wave Formate

前言

  • Wave文件应该是音频领域使用最广泛的容器格式了.对于Wave文件的解析器(Parser),也是构建音频引擎的重要组件.本文受数字音频规范与程序设计与Waveless项目启发,意在搭建一个简单parser模型来更好的理解Wave文件格式.

一些Wave格式标准

  • Wave格式是属于RIFF(Resourse Interchange File Format)规范的格式之一.这个规范下的多媒体格式还有.RMI/.RMN/.ANI

  • Wave文件支持所有ACM规范下的编码格式,常用的有PCM/ADPCM.

  • Wave文件常用采样率为11.025kHz/44.1kHz/48kHz/96kHz.常用采样精度为8bit/16bit/24bit/32bit

  • 8bit采样精度使用unsigned char来表示,取值范围0~255.16bit采样精度使用signed short来表示,取值范围-32768 ~ 32767.24bit/32bit通常采用浮点类型来表示,也可以使用signed int来表示,取值范围−8388608 ~ 8388607.

  • PCM编码(非ADPCM压缩编码)的wav文件大小 = 采样率(sample rate) * 采样精度(bit depth) * 声道数 * 时长/8

文件结构

  • wave文件由基础Chunk和SubChunk组成，下图是基础Chunk。包含了基本文件信息和采样点数据。

  • 根据8~12字节的format信息可以将文件类型分为三种

  1. PCM Formate (WAVE_FORMAT_PCM)

    • 标准的PCM文件类型

  2. Non-PCM Formats (WAVE_FORMAT_IEEE_FLOAT / WAVE_FORMAT_ALAW / WAVE_FORMAT_MULAW)

    • 浮点编码或者压缩编码的PCM类型

  3. Extensible Format (WAVE_FORMAT_EXTENSIBLE)

    • fmt chunk中有其他sub-chunk时的类型

Chunk数据

  • Wave格式的元数据是通过各类Chunk数据块组合而成,不同类型Chunk块储存不同的种类元数据,Chunk也是元数据组合的最小单位.

    • RIFF Chunk

      struct RIFFChunk
  {
    //RIFF规范标识符 : "RIFF"
    char ckID[4];
    //RIFF Chunk 大小
    unsigned long ckSize;
    //Wave格式ID: "WAVE"
    char WavID[4];
  };

    • Standard Format Chunk

    struct StandardPCMFmtChunk
{
  //Chunk ID: "fmt"
  char ckID[4];
  //fmt Chunk 大小
  unsigned long ckSize;
  //音频数据的编码格式
  unsigned short wFormatTag;
  //声道数
  unsigned short nChannels;
  //采样率
  unsigned long nSamplesPerSec;
  //每秒码率
  unsigned long nAvgBytesPerSec;
  //对齐占位数据 2=16-bit mono, 4=16-bit stereo
  unsigned short nBlockAlign;
  //采样精度
  unsigned short wBitsPerSample;
};

    • Non-PCM Format Chunk

    struct NonPcmFmtChunk
{
  //与Standard Format Chunk相同
  StandardPCMFmtChunk standardFmtChunk;
    //extension Chunk的大小
    unsigned short cbSize;
};

    • Extensible Format Chunk

    struct ExtensibleFmtChunk
{
  //与Standard Format Chunk相同
    StandardPCMFmtChunk standardFmtChunk;
    //extension Chunk的大小
    unsigned short cbSize;
    //实际有效的编码位,比如12bit的编码方式采样精度显示的是16bit,实际有效编码位只有12bit
    unsigned short wValidBitsPerSample;
    //回放音响的布局
    unsigned long dwChannelMask;
    //GUID标识符(头两位是subformat信息,剩余12位是GUID标识符)
    char SubFormat[16];
};

    • Data Chunk

    struct DataChunk
{
    //Chunk ID: "data"
	char ckID[4];
    //实际音频数据大小
    unsigned long ckSize;
};

    • Fact Chunk

    struct FactChunk
{
    //Chunk ID: "fact"
    char ckID[4];
    //fact Chunk 大小
    unsigned long ckSize;
    //每个声道采样数
    unsigned long dwSampleLength;
};

    • Bext Chunk

    struct BextChunk
{
	//Chunk ID = "bext"
	char ckID[4];
  //bext chunk 大小
	unsigned long ckSize;
  //文件描述信息
	char Description[256];
  //创作者信息
	char Originator[32];
  //创作者关联信息
	char OriginatorReference[32];
  //创作日期
	char OriginationDate[10];
  //创作时间 <<hh::mm::ss>>
	char OriginationTime[8];
	//对齐占位数据
	unsigned short Align = 0;
  //参考时间 低位
	unsigned long TimeReferenceLow;
  //参考时间 高位
	unsigned long TimeReferenceHigh;
  //BWF版本号
	unsigned short Version;
  //SMPTE UMID数据
	char UMID[64];
  //集成响度值 LUFS
	short LoudnessValue;
  //响度范围值 LU
	short LoudnessRange;
  //TruePeak峰值 dBTP
	short MaxTruePeakLevel;
  //瞬时响度均值 LUFS
	short MaxMomentaryLoudness;
  //ShortTerm响度均值 LUFS
	short MaxShortTermLoudness;
  //预留空间 现在全为0
	char Reserved[180];
};

    • Junk Chunk

    struct JunkChunk
{
	//Chunk ID = "JUNK"
	char ckID[4];
  //Junk Chunk 大小
	unsigned long ckSize;
  //无效数据
	char initialData[74];
};

    • RF64 Chunk

    struct RF64Chunk
{
	// Chunk ID = "RF64"
	char ckID[4]; 
	// -1 不要使用这个数据,用DS64 Chunk里面的riffSizeHigh 和 riffSizeLow来获取大小
	unsigned long ckSize;
	// Type ID = "WAVE"
	char rf64Type[4]; 
};

    • DS64 Chunk

    struct DS64Chunk
{
	// Chunk ID = "DS64"
	char ckID[4];
  //DS64 Chunk大小
	unsigned long ckSize;
  //RF64 Chunk大小的低四位
	unsigned long RIFFSizeLow;
  //RF64 Chunk大小的高四位
	unsigned long RIFFSizeHigh;
  //Data Chunk数据大小的低四位
	unsigned long DataSizeLow;
  //Data Chunk数据大小的高四位
	unsigned long DataSizeHigh;
  //fact Chunk采样数大小低四位
	unsigned long SampleCountLow;
  //fact Chunk采样数大小高四位
	unsigned long SampleCountHigh;
  //Table数组的大小
	unsigned long TableLength;
};

    • PAD Chunk

    struct PadChunk
{
  //Chunk ID = "PAD"
	char ckID[4];
  //Pad Chunk大小
	unsigned long ckSize;
};

    • 一些不常用扩展Chunk

    1. Cue Chunk
    2. Playlist Chunk
    3. Associated Data Chunk
    4. Instrument Chunk
    5. Sample Chunk

Header分类

  • Standard PCM

struct StandardWavHead
{
	RIFFChunk RIFFChunk;
	StandardPCMFmtChunk fmtChunk;
	DataChunk DataChunk;
};

  • Non-PCM

//基本少有单独是Non-PCM类型的Header,因为Non-PCM Header只是包含了Extensible  Header的基础数据类型
struct NonPcmHead
{
	RIFFChunk RIFFChunk;
	NonPcmFmtChunk fmtChunk;
	FactChunk FactChunk;
	DataChunk DataChunk;
};

  • Extensible

struct ExtensibleWavHead
{
	RIFFChunk RIFFChunk;
	ExtensibleFmtChunk fmtChunk;
	FactChunk FactChunk;
	DataChunk DataChunk;
};

struct ExtensibleNoFactHead
{
	RIFFChunk RIFFChunk;
	ExtensibleFmtChunk fmtChunk;
	DataChunk DataChunk;
};

  • BWF

//EBU制定的Wave格式扩展标准,增加了大量的元数据供使用
struct StandardBWFHead
{
	RIFFChunk RIFFChunk;
	StandardPCMFmtChunk fmtChunk;
	BextChunk BextChunk;
	JunkChunk JunkChunk;
	DataChunk DataChunk;
};

struct RF64Header
{
	RF64Chunk RF64Chunk;
	DS64Chunk DS64Chunk;
	//std::vector<CS64Chunk> TableChunk;
	ExtensibleFmtChunk fmtChunk;
	BextChunk BextChunk;
	//JunkChunk JunkChunk;
	DataChunk DataChunk;
};

struct ExtensibleBWFHead
{
	RIFFChunk RIFFChunk;
	ExtensibleFmtChunk fmtChunk;
	FactChunk FactChunk;
	BextChunk BextChunk;
	JunkChunk JunkChunk;
	DataChunk DataChunk;
};

Data数据类型

enum class WavDataType
{
	Invalid,
	//usigned int8
	WavData8bit,
	//int12
	WavData12bit,
	//int16
	WavData16bit,
	//int24
	WavData24bit,
	//float32
	WavData32bit,
	//float64
	WavData64bit
};
signed-integer
PCM data stored as signed (‘two’s complement’) integers. Commonly used with a 1or 24 −bit encoding size. A value of 0 represents minimum signal power.

unsigned-integer
PCM data stored as unsigned integers. Commonly used with an 8-bit encoding size. value of 0 represents maximum signal power.

floating-point
PCM data stored as IEEE 753 single precision (32-bit) or double precision (64-bitfloating-point (‘real’) numbers. A value of 0 represents minimum signal power.

a-law 
International telephony standard for logarithmic encoding to 8 bits per sample. Ihas a precision equivalent to roughly 13-bit PCM and is sometimes encoded witreversed bitordering (see the −X option).

u-law, mu-law
North American telephony standard for logarithmic encoding to 8 bits per sampleA.k.a.µ-law. It has a precision equivalent to roughly 14-bit PCM and is sometimeencoded with reversed bit-ordering (see the −X option).

oki-adpcm
OKI (a.k.a. VOX, Dialogic, or Intel) 4-bit ADPCM; it has a precision equivalent to
roughly 12-bit PCM. ADPCM is a form of audio compression that has a goocompromise between audio quality and encoding/decoding speed.

ima-adpcm
IMA (a.k.a. DVI) 4-bit ADPCM; it has a precision equivalent to roughly 13-bit PCM.

ms-adpcm
Microsoft 4-bit ADPCM; it has a precision equivalent to roughly 14-bit PCM.

gsm-full-rate
GSM is currently used for the vast majority of the world’s digital wirelestelephone calls.It utilises several audio formats with different bit-rates anassociated speech quality.

  • Data数据的一些标准

    • I/O

      Wave采样数据点的取值范围和采样精度相关联.CPU对于数据帧的IO操作是以8位为基础单位的,所以数模转换器产生(ADC)产生的数据也是8的整数倍.

    • 左对齐(left-justified)

      采样点数据使用左对齐,剩余的比特位在右边用零填充.比如12bit数据101011101011补全为1010111010110000.

    • 小端编码(little endian)

      低比特位在低地址位.内存内布局: 10110000|10101110.

    • 采样帧(sample frame)

      • 同时播放的采样点称作采样帧。多声道文件的采样点数据是交替储存的。
      • stereo [left | right]
      • 3.0 [left | right | center]
      • quad [front left | front right | rear left | rear right]
      • 4.0 [left | center | right | surround]
      • 5.1 [left center | left | center | right center | right | surround]

WaveParser

GitHub

  • Parser设计思路

    我准备了22个各种Header类型的测试文件,以正确解析所有文件为标准,只负责Header数据的解析,暂时没有Header信息的编辑以及Data部分的处理.

  • Parser的使用

    #include "WavParser.h"
#include "WavUtility.h"

int main()
{
  const std::string& FilePath = "MyFileFolder/MyFile.wav"

  auto MyParser = new WavParser();
  if(MyParser->LoadFile(FilePath))
  {
    MyParser->Parser();
  }

  std::cin.get();
  delete MyParser;
  return 0;
}

  • Parser的测试文件

    enum class TestFileType
{
  //WAVE file, stereo unsigned 8-bit data
    Uint8Stereo,
  //WAVE file, stereo A-law data
    Uint8ALawStereo,
  //WAVE file, stereo µ-law data
    UInt8MuLawStereo,
  //WAVE file, stereo 12/16-bit data
    Int12Stereo,
  //WAVE file, stereo 16-bit data
    Int16Stereo,
  //WAVE file, stereo 24-bit data
    Int24Stereo,
  //WAVE file, stereo 32-bit data
    Int32Stereo,
  //WAVE file, stereo 32-bit float data
    Float32Stereo,
  //WAVE file, stereo 64-bit float data
    Float64Stereo,
  //WAVE (WAVE_FORMAT_EXTENSIBLE) file, stereo unsigned 8-bit data
    Uint8StereoExt,
  //WAVE (WAVE_FORMAT_EXTENSIBLE) file, stereo A-law data
    UInt8ALawStereoExt,
  //WAVE (WAVE_FORMAT_EXTENSIBLE) file, stereo µ-law data
    UInt8MuLawStereoExt,
  //WAVE (WAVE_FORMAT_EXTENSIBLE) file, stereo 12/16-bit data
    Int12StereoExt,
  //WAVE (WAVE_FORMAT_EXTENSIBLE) file, stereo 16-bit data
    Int16StereoExt,
  //WAVE (WAVE_FORMAT_EXTENSIBLE) file, stereo 24-bit data
    Int24StereoExt,
  //WAVE (WAVE_FORMAT_EXTENSIBLE) file, stereo 32-bit data
    Int32StereoExt,
  //WAVE (WAVE_FORMAT_EXTENSIBLE) file, stereo 32-bit float data
    Float32StereoExt,
  //WAVE (WAVE_FORMAT_EXTENSIBLE) file, stereo 64-bit float data
    Float64StereoExt,
  //6-channel WAVE file with speaker locations FL FR FC LF BL BR, 44100 Hz, 16-bit
    Int16Channel_6,
  //8-channel WAVE file with speaker locations FL FR FC LF BL BR - -, 48000 Hz, 24-bit
    Int24Channel_8,
  //WAVE file, BWF Type, 16-bit data
    Int16Bwf,
  //WAVE file, RF64 Type, 24-bit Data
    Int24RF64
};

int main()
{
  const std::string& FilePath = GetTestFilePath(TestFileType::Int16Stereo);

  ......

}