3.B:vendor opencore-amr + AMRCodec Swift wrapper,打通 AMR↔WAV 编解码

为支撑 H5 的 prepareaudio / mediaTypeAudio 录音上传 + 远端语音播放
(docs/H5-Native-Contract.md §3.1 [4][5] + §3.2 [9])接入 AMR-NB
编解码能力。AMR-NB 是跨端(iOS / Android / 旧版本客户端)契约硬约束,
不可换 AAC / Opus。

## 接入路径

直接 vendor daoqi msext 现成的 fat .a(已上线多年),不打 xcframework /
不重编源码:
- Vendor/OpenCoreAMR/v0.1.x/libopencore-amrnb.a:daoqi 同款 5-slice fat
  库(含 2026-06-20 lipo -segalign 8 修过的对齐版本)
- Vendor/OpenCoreAMR/v0.1.x/{interf_dec,interf_enc}.h:amrnb 公开头
- Vendor/OpenCoreAMR/README.md:来源、版本、修复历史、为何不打 xcframework
- ylgamehall/Source/Audio/AMRFileCodec/{amrFileCodec.h,amrFileCodec.mm}:
  daoqi 同款 wrapper(文件级 WAV↔AMR 处理)
- ylgamehall/Source/Audio/AMRCodec.swift:Swift 薄 wrapper(wavToAmr /
  amrToWav 两个纯函数 + AMRCodec.Error)

理由见 Vendor/OpenCoreAMR/README.md「为何不打 xcframework / 不重编源码」
一节,本质是 opencore-amr 整个 iOS 生态没有"现代主流"分发形态,社区无
官方 SPM / xcframework 维护,强行打反而引入新风险。

## 关键技术点:extern "C" 修复 ObjC++ name mangling

amrFileCodec.h 在原项目里没有 extern "C" 包裹函数声明,daoqi 能跑是
因为调用方 VoiceConverter.mm 也是 ObjC++,双方按相同 C++ name mangling
链接能对上。

新外壳从 Swift 通过 bridging-header 调用,Swift 按 C 函数解析,期望符号
_EncodeWAVEFileToAMRFile / _DecodeWAVEFileToWAVEFile(C 链接);但
.mm 编译时默认 C++ name mangling,符号变成 __Z25Encode...,导致链接器
报 Undefined symbol。

修复:在 amrFileCodec.h 的函数声明外加 #ifdef __cplusplus / extern "C"
包裹。这是标准 C/C++ 头文件兼容写法,对 daoqi 老路径无影响(ObjC++
调用方仍能链接 C-linkage 函数)。

## .gitignore:白名单 Vendor 下的 .a

发现历史遗留问题:原 .gitignore *.a 通配规则把 libWeChatSDK.a 也排除
了,意味着别人 clone 此仓库后链接失败。本次顺带修复:

  !Vendor/**/*.a

让 Vendor 下的 .a 显式纳入跟踪。本提交同时把 libWeChatSDK.a(历史
缺失)+ libopencore-amrnb.a(本次新增)一并加入 git 跟踪。

## 工程改动

pbxproj 改动由用户在 Xcode UI 完成(4 项 Build Settings:Add .a 到
Frameworks group / Header Search Paths +2 / Library Search Paths +1;
EXCLUDED_ARCHS arm64-simulator 因真机调试场景暂未配,需要时再加)。

## 验证

BuildProject 通过,0 错误,符号链接正确。XcodeRefreshCodeIssuesInFile
对 AMRCodec.swift 0 诊断。
This commit is contained in:
joywayer
2026-06-24 01:20:30 +08:00
parent 5463b07fb6
commit 5b38456c66
11 changed files with 776 additions and 6 deletions
+69
View File
@@ -0,0 +1,69 @@
//
// AMRCodec.swift
// ylgamehall
//
// AMR-NB WAV Swift wrapper
//
// ##
//
// AMR-NB H5 iOS / Android /
// AMR-NB audiourl Vendor/OpenCoreAMR/README.md
//
// daoqi 沿 `amrFileCodec.mm`C++ WAV
// header PCM opencore-amr AMR / Swift
//
//
// - Swift amrFileCodec C byte FILE* /
// memcpy / Swift daoqi
// 线 N
// - amrFileCodec.{h,mm} Source/Audio/AMRFileCodec/ synchronized
// group amrFileCodec.h ylgamehall-Bridging-Header.h
// Swift
//
// ##
//
// - prepareaudio`wavToAmr(_:to:)` AVAudioRecorder wav
// amr
// - mediaTypeAudio`amrToWav(_:to:)` amr wav
// AVAudioPlayer
//
// docs/H5-Native-Contract.md §3.1 45 + §3.2 9
//
import Foundation
public enum AMRCodec {
public enum Error: Swift.Error, Sendable {
/// EncodeWAVEFileToAMRFile 0fopen 0
case wavToAmrFailed(wav: String, amr: String)
/// DecodeAMRFileToWAVEFile 0amr / / wav
case amrToWavFailed(amr: String, wav: String)
}
/// WAV AMR
///
/// wav 8kHz / 16bit / monoVoiceRecorderBaseVC.m
/// `getAudioRecorderSettingDict` msext
/// amrFileCodec ReadPCMFrame 160 sample 0
///
/// - Throws: `AMRCodec.Error.wavToAmrFailed` C 0
public static func wavToAmr(_ wav: URL, to amr: URL) throws {
let frames = EncodeWAVEFileToAMRFile(wav.path, amr.path, 1, 16)
guard frames > 0 else {
throw Error.wavToAmrFailed(wav: wav.path, amr: amr.path)
}
}
/// AMR WAV
///
/// wav 8kHz / 16bit / mono AVAudioPlayer
///
/// - Throws: `AMRCodec.Error.amrToWavFailed` C 0
public static func amrToWav(_ amr: URL, to wav: URL) throws {
let frames = DecodeAMRFileToWAVEFile(amr.path, wav.path)
guard frames > 0 else {
throw Error.amrToWavFailed(amr: amr.path, wav: wav.path)
}
}
}
+90
View File
@@ -0,0 +1,90 @@
//
// amrFileCodec.h
// amrDemoForiOS
//
// Created by Tang Xiaoping on 9/27/11.
// Copyright 2011 test. All rights reserved.
//
#ifndef amrFileCodec_h
#define amrFileCodec_h
#include <stdlib.h>
#include <string.h>
#include <stdio.h>
#include "interf_dec.h"
#include "interf_enc.h"
#define AMR_MAGIC_NUMBER "#!AMR\n"
#define PCM_FRAME_SIZE 160 // 8khz 8000*0.02=160
#define MAX_AMR_FRAME_SIZE 32
#define AMR_FRAME_COUNT_PER_SECOND 50
typedef struct
{
char chChunkID[4];
int nChunkSize;
}XCHUNKHEADER;
typedef struct
{
short nFormatTag;
short nChannels;
int nSamplesPerSec;
int nAvgBytesPerSec;
short nBlockAlign;
short nBitsPerSample;
}WAVEFORMAT;
typedef struct
{
short nFormatTag;
short nChannels;
int nSamplesPerSec;
int nAvgBytesPerSec;
short nBlockAlign;
short nBitsPerSample;
short nExSize;
}WAVEFORMATX;
typedef struct
{
char chRiffID[4];
int nRiffSize;
char chRiffFormat[4];
}RIFFHEADER;
typedef struct
{
char chFmtID[4];
int nFmtSize;
WAVEFORMAT wf;
}FMTBLOCK;
// extern "C" 在 ObjC++ (.mm) 编译时禁用 C++ name mangling
// 保证 .o 里函数符号是 _EncodeWAVEFileToAMRFile / _DecodeAMRFileToWAVEFileC 链接),
// 与 Swift bridging-header 从 amrFileCodec.h 推导出的 C 函数签名匹配。
//
// daoqi msext 原版无 extern "C" 包裹,因为调用方 VoiceConverter.mm 也是 ObjC++
// 双方都按 C++ name mangling 链接能匹配上。新外壳从 Swift 通过 bridging-header
// 调用,Swift 按 C 函数解析,必须显式标 C linkage。
#ifdef __cplusplus
extern "C" {
#endif
// WAVE音频采样频率是8khz
// 音频样本单元数 = 8000*0.02 = 160 (由采样频率决定)
// 声道数 1 : 160
// 2 : 160*2 = 320
// bps决定样本(sample)大小
// bps = 8 --> 8位 unsigned char
// 16 --> 16位 unsigned short
int EncodeWAVEFileToAMRFile(const char* pchWAVEFilename, const char* pchAMRFileName, int nChannels, int nBitsPerSample);
// 将AMR文件解码成WAVE文件
int DecodeAMRFileToWAVEFile(const char* pchAMRFileName, const char* pchWAVEFilename);
#ifdef __cplusplus
}
#endif
#endif
+393
View File
@@ -0,0 +1,393 @@
//
// amrFileCodec.cpp
// amrDemoForiOS
//
// Created by Tang Xiaoping on 9/27/11.
// Copyright 2011 test. All rights reserved.
//
#include "amrFileCodec.h"
int amrEncodeMode[] = {4750, 5150, 5900, 6700, 7400, 7950, 10200, 12200}; // amr 编码方式
// 从WAVE文件中跳过WAVE文件头,直接到PCM音频数据
void SkipToPCMAudioData(FILE* fpwave)
{
RIFFHEADER riff;
FMTBLOCK fmt;
XCHUNKHEADER chunk;
WAVEFORMATX wfx;
int bDataBlock = 0;
// 1. 读RIFF头
fread(&riff, 1, sizeof(RIFFHEADER), fpwave);
// 2. 读FMT块 - 如果 fmt.nFmtSize>16 说明需要还有一个附属大小没有读
fread(&chunk, 1, sizeof(XCHUNKHEADER), fpwave);
if ( chunk.nChunkSize>16 )
{
fread(&wfx, 1, sizeof(WAVEFORMATX), fpwave);
}
else
{
memcpy(fmt.chFmtID, chunk.chChunkID, 4);
fmt.nFmtSize = chunk.nChunkSize;
fread(&fmt.wf, 1, sizeof(WAVEFORMAT), fpwave);
}
// 3.转到data块 - 有些还有fact块等。
while(!bDataBlock)
{
fread(&chunk, 1, sizeof(XCHUNKHEADER), fpwave);
if ( !memcmp(chunk.chChunkID, "data", 4) )
{
bDataBlock = 1;
break;
}
// 因为这个不是data块,就跳过块数据
fseek(fpwave, chunk.nChunkSize, SEEK_CUR);
}
}
// 从WAVE文件读一个完整的PCM音频帧
// 返回值: 0-错误 >0: 完整帧大小
int ReadPCMFrame(short speech[], FILE* fpwave, int nChannels, int nBitsPerSample)
{
int nRead = 0;
int x = 0, y=0;
// unsigned short ush1=0, ush2=0, ush=0;
// 原始PCM音频帧数据
unsigned char pcmFrame_8b1[PCM_FRAME_SIZE];
unsigned char pcmFrame_8b2[PCM_FRAME_SIZE<<1];
unsigned short pcmFrame_16b1[PCM_FRAME_SIZE];
unsigned short pcmFrame_16b2[PCM_FRAME_SIZE<<1];
if (nBitsPerSample==8 && nChannels==1)
{
nRead = fread(pcmFrame_8b1, (nBitsPerSample/8), PCM_FRAME_SIZE*nChannels, fpwave);
for(x=0; x<PCM_FRAME_SIZE; x++)
{
speech[x] =(short)((short)pcmFrame_8b1[x] << 7);
}
}
else
if (nBitsPerSample==8 && nChannels==2)
{
nRead = fread(pcmFrame_8b2, (nBitsPerSample/8), PCM_FRAME_SIZE*nChannels, fpwave);
for( x=0, y=0; y<PCM_FRAME_SIZE; y++,x+=2 )
{
// 1 - 取两个声道之左声道
speech[y] =(short)((short)pcmFrame_8b2[x+0] << 7);
// 2 - 取两个声道之右声道
//speech[y] =(short)((short)pcmFrame_8b2[x+1] << 7);
// 3 - 取两个声道的平均值
//ush1 = (short)pcmFrame_8b2[x+0];
//ush2 = (short)pcmFrame_8b2[x+1];
//ush = (ush1 + ush2) >> 1;
//speech[y] = (short)((short)ush << 7);
}
}
else
if (nBitsPerSample==16 && nChannels==1)
{
nRead = fread(pcmFrame_16b1, (nBitsPerSample/8), PCM_FRAME_SIZE*nChannels, fpwave);
for(x=0; x<PCM_FRAME_SIZE; x++)
{
speech[x] = (short)pcmFrame_16b1[x+0];
}
}
else
if (nBitsPerSample==16 && nChannels==2)
{
nRead = fread(pcmFrame_16b2, (nBitsPerSample/8), PCM_FRAME_SIZE*nChannels, fpwave);
for( x=0, y=0; y<PCM_FRAME_SIZE; y++,x+=2 )
{
//speech[y] = (short)pcmFrame_16b2[x+0];
speech[y] = (short)((int)((int)pcmFrame_16b2[x+0] + (int)pcmFrame_16b2[x+1])) >> 1;
}
}
// 如果读到的数据不是一个完整的PCM帧, 就返回0
if (nRead<PCM_FRAME_SIZE*nChannels) return 0;
return nRead;
}
// WAVE音频采样频率是8khz
// 音频样本单元数 = 8000*0.02 = 160 (由采样频率决定)
// 声道数 1 : 160
// 2 : 160*2 = 320
// bps决定样本(sample)大小
// bps = 8 --> 8位 unsigned char
// 16 --> 16位 unsigned short
int EncodeWAVEFileToAMRFile(const char* pchWAVEFilename, const char* pchAMRFileName, int nChannels, int nBitsPerSample)
{
FILE* fpwave;
FILE* fpamr;
/* input speech vector */
short speech[160];
/* counters */
int byte_counter, frames = 0, bytes = 0;
/* pointer to encoder state structure */
void *enstate;
/* requested mode */
enum Mode req_mode = MR122;
int dtx = 0;
/* bitstream filetype */
unsigned char amrFrame[MAX_AMR_FRAME_SIZE];
fpwave = fopen(pchWAVEFilename, "rb");
if (fpwave == NULL)
{
return 0;
}
// 创建并初始化amr文件
fpamr = fopen(pchAMRFileName, "wb");
if (fpamr == NULL)
{
fclose(fpwave);
return 0;
}
/* write magic number to indicate single channel AMR file storage format */
bytes = fwrite(AMR_MAGIC_NUMBER, sizeof(char), strlen(AMR_MAGIC_NUMBER), fpamr);
/* skip to pcm audio data*/
SkipToPCMAudioData(fpwave);
enstate = Encoder_Interface_init(dtx);
while(1)
{
// read one pcm frame
if (!ReadPCMFrame(speech, fpwave, nChannels, nBitsPerSample)) break;
frames++;
/* call encoder */
byte_counter = Encoder_Interface_Encode(enstate, req_mode, speech, amrFrame, 0);
bytes += byte_counter;
fwrite(amrFrame, sizeof (unsigned char), byte_counter, fpamr );
}
Encoder_Interface_exit(enstate);
fclose(fpamr);
fclose(fpwave);
return frames;
}
#pragma mark - Decode
//decode
void WriteWAVEFileHeader(FILE* fpwave, int nFrame)
{
char tag[10] = "";
// 1. 写RIFF头
RIFFHEADER riff;
strcpy(tag, "RIFF");
memcpy(riff.chRiffID, tag, 4);
riff.nRiffSize = 4 // WAVE
+ sizeof(XCHUNKHEADER) // fmt
+ sizeof(WAVEFORMATX) // WAVEFORMATX
+ sizeof(XCHUNKHEADER) // DATA
+ nFrame*160*sizeof(short); //
strcpy(tag, "WAVE");
memcpy(riff.chRiffFormat, tag, 4);
fwrite(&riff, 1, sizeof(RIFFHEADER), fpwave);
// 2. 写FMT块
XCHUNKHEADER chunk;
WAVEFORMATX wfx;
strcpy(tag, "fmt ");
memcpy(chunk.chChunkID, tag, 4);
chunk.nChunkSize = sizeof(WAVEFORMATX);
fwrite(&chunk, 1, sizeof(XCHUNKHEADER), fpwave);
memset(&wfx, 0, sizeof(WAVEFORMATX));
wfx.nFormatTag = 1;
wfx.nChannels = 1; // 单声道
wfx.nSamplesPerSec = 8000; // 8khz
wfx.nAvgBytesPerSec = 16000;
wfx.nBlockAlign = 2;
wfx.nBitsPerSample = 16; // 16位
fwrite(&wfx, 1, sizeof(WAVEFORMATX), fpwave);
// 3. 写data块头
strcpy(tag, "data");
memcpy(chunk.chChunkID, tag, 4);
chunk.nChunkSize = nFrame*160*sizeof(short);
fwrite(&chunk, 1, sizeof(XCHUNKHEADER), fpwave);
}
const int myround(const double x)
{
return((int)(x+0.5));
}
// 根据帧头计算当前帧大小
int caclAMRFrameSize(unsigned char frameHeader)
{
int mode;
int temp1 = 0;
int temp2 = 0;
int frameSize;
temp1 = frameHeader;
// 编码方式编号 = 帧头的3-6位
temp1 &= 0x78; // 0111-1000
temp1 >>= 3;
mode = amrEncodeMode[temp1];
// 计算amr音频数据帧大小
// 原理: amr 一帧对应20ms,那么一秒有50帧的音频数据
temp2 = myround((double)(((double)mode / (double)AMR_FRAME_COUNT_PER_SECOND) / (double)8));
frameSize = myround((double)temp2 + 0.5);
return frameSize;
}
// 读第一个帧 - (参考帧)
// 返回值: 0-出错; 1-正确
int ReadAMRFrameFirst(FILE* fpamr, unsigned char frameBuffer[], int* stdFrameSize, unsigned char* stdFrameHeader)
{
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wsizeof-array-argument"
#pragma clang diagnostic ignored "-Wsizeof-pointer-memaccess"
memset(frameBuffer, 0, sizeof(frameBuffer));
#pragma clang diagnostic pop
// 先读帧头
fread(stdFrameHeader, 1, sizeof(unsigned char), fpamr);
if (feof(fpamr)) return 0;
// 根据帧头计算帧大小
*stdFrameSize = caclAMRFrameSize(*stdFrameHeader);
// 读首帧
frameBuffer[0] = *stdFrameHeader;
fread(&(frameBuffer[1]), 1, (*stdFrameSize-1)*sizeof(unsigned char), fpamr);
if (feof(fpamr)) return 0;
return 1;
}
// 返回值: 0-出错; 1-正确
int ReadAMRFrame(FILE* fpamr, unsigned char frameBuffer[], int stdFrameSize, unsigned char stdFrameHeader)
{
int bytes = 0;
unsigned char frameHeader; // 帧头
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wsizeof-array-argument"
#pragma clang diagnostic ignored "-Wsizeof-pointer-memaccess"
memset(frameBuffer, 0, sizeof(frameBuffer));
#pragma clang diagnostic pop
// 读帧头
// 如果是坏帧(不是标准帧头),则继续读下一个字节,直到读到标准帧头
while(1)
{
bytes = fread(&frameHeader, 1, sizeof(unsigned char), fpamr);
if (feof(fpamr)) return 0;
if (frameHeader == stdFrameHeader) break;
}
// 读该帧的语音数据(帧头已经读过)
frameBuffer[0] = frameHeader;
bytes = fread(&(frameBuffer[1]), 1, (stdFrameSize-1)*sizeof(unsigned char), fpamr);
if (feof(fpamr)) return 0;
return 1;
}
// 将AMR文件解码成WAVE文件
int DecodeAMRFileToWAVEFile(const char* pchAMRFileName, const char* pchWAVEFilename)
{
FILE* fpamr = NULL;
FILE* fpwave = NULL;
char magic[8];
void * destate;
int nFrameCount = 0;
int stdFrameSize;
unsigned char stdFrameHeader;
unsigned char amrFrame[MAX_AMR_FRAME_SIZE];
short pcmFrame[PCM_FRAME_SIZE];
// NSString * path = [[NSBundle mainBundle] pathForResource: @"test" ofType: @"amr"];
// fpamr = fopen([path cStringUsingEncoding:NSASCIIStringEncoding], "rb");
fpamr = fopen(pchAMRFileName, "rb");
if ( fpamr==NULL ) return 0;
// 检查amr文件头
fread(magic, sizeof(char), strlen(AMR_MAGIC_NUMBER), fpamr);
if (strncmp(magic, AMR_MAGIC_NUMBER, strlen(AMR_MAGIC_NUMBER)))
{
fclose(fpamr);
return 0;
}
// 创建并初始化WAVE文件
// NSArray *paths = NSSearchPathForDirectoriesInDomains(NSDocumentDirectory, NSUserDomainMask, YES);
// NSString *documentPath = [paths objectAtIndex:0];
// NSString *docFilePath = [documentPath stringByAppendingPathComponent:[NSString stringWithFormat:@"%s", pchWAVEFilename]];
// NSLog(@"documentPath=%@", documentPath);
//
// fpwave = fopen([docFilePath cStringUsingEncoding:NSASCIIStringEncoding], "wb");
fpwave = fopen(pchWAVEFilename,"wb");
WriteWAVEFileHeader(fpwave, nFrameCount);
/* init decoder */
destate = Decoder_Interface_init();
// 读第一帧 - 作为参考帧
memset(amrFrame, 0, sizeof(amrFrame));
memset(pcmFrame, 0, sizeof(pcmFrame));
ReadAMRFrameFirst(fpamr, amrFrame, &stdFrameSize, &stdFrameHeader);
// 解码一个AMR音频帧成PCM数据
Decoder_Interface_Decode(destate, amrFrame, pcmFrame, 0);
nFrameCount++;
fwrite(pcmFrame, sizeof(short), PCM_FRAME_SIZE, fpwave);
// 逐帧解码AMR并写到WAVE文件里
while(1)
{
memset(amrFrame, 0, sizeof(amrFrame));
memset(pcmFrame, 0, sizeof(pcmFrame));
if (!ReadAMRFrame(fpamr, amrFrame, stdFrameSize, stdFrameHeader)) break;
// 解码一个AMR音频帧成PCM数据 (8k-16b-单声道)
Decoder_Interface_Decode(destate, amrFrame, pcmFrame, 0);
nFrameCount++;
fwrite(pcmFrame, sizeof(short), PCM_FRAME_SIZE, fpwave);
}
printf("frame = %d\n", nFrameCount);
Decoder_Interface_exit(destate);
fclose(fpwave);
// 重写WAVE文件头
// fpwave = fopen([docFilePath cStringUsingEncoding:NSASCIIStringEncoding], "r+");
fpwave = fopen(pchWAVEFilename, "r+");
WriteWAVEFileHeader(fpwave, nFrameCount);
fclose(fpwave);
return nFrameCount;
}