iOS音频编程之实时语音通信
2016-08-01 本文已影响2323人
JustinYang
title: iOS音频编程之实时语音通信
date: 2016-07-14
tags: AAC Converter,Audio Queue,Audio Unit,MultipeerConnectivity
博客地址
iOS音频编程之实时语音通信
需求：手机通过Mic采集PCM编码的原始音频数据，将PCM转换为AAC编码格式，通过MultipeerConnectivity框架连接手机并发送AAC数据，在接收端使用Audio Queue播放收到的AAC音频
音频设置
对音频以44.1KHZ的采样率来采样,以64000的比特率对PCM进行AAC转码
1）对AVAudioSession的设置
NSError *error;
self.session = [AVAudioSession sharedInstance];
[self.session setCategory:AVAudioSessionCategoryPlayAndRecord error:&error];
handleError(error);
//route变化监听
[[NSNotificationCenter defaultCenter] addObserver:self selector:@selector(audioSessionRouteChangeHandle:) name:AVAudioSessionRouteChangeNotification object:self.session];
[self.session setPreferredIOBufferDuration:0.005 error:&error];
handleError(error);
[self.session setPreferredSampleRate:kSmaple error:&error];
handleError(error);
//[self.session overrideOutputAudioPort:AVAudioSessionPortOverrideSpeaker error:&error];
//handleError(error);
[self.session setActive:YES error:&error];
handleError(error);
-(void)audioSessionRouteChangeHandle:(NSNotification *)noti{
// NSError *error;
// [self.session overrideOutputAudioPort:AVAudioSessionPortOverrideSpeaker error:&error];
// handleError(error);
[self.session setActive:YES error:nil];
if (self.startRecord) {
CheckError(AudioOutputUnitStart(_toneUnit), "couldnt start audio unit");
}
}
音频输入输出路径改变会触发audioSessionRouteChangeHandle,如果想一直让音频从手机的扬声器输出需要在每次Route改变时，把音频输出重定向到AVAudioSessionPortOverrideSpeaker,否则为手机听筒输出音频;其他设置说明请参照iOS音频编程之变声处理的初始化部分
2)对Audio Unit的设置
AudioComponentDescription acd;
acd.componentType = kAudioUnitType_Output;
acd.componentSubType = kAudioUnitSubType_RemoteIO;
acd.componentFlags = 0;
acd.componentFlagsMask = 0;
acd.componentManufacturer = kAudioUnitManufacturer_Apple;
AudioComponent inputComponent = AudioComponentFindNext(NULL, &acd);
AudioComponentInstanceNew(inputComponent, &_toneUnit);
UInt32 enable = 1;
AudioUnitSetProperty(_toneUnit,
kAudioOutputUnitProperty_EnableIO,
kAudioUnitScope_Input,
kInputBus,
&enable,
sizeof(enable));
mAudioFormat.mSampleRate = kSmaple;//采样率
mAudioFormat.mFormatID = kAudioFormatLinearPCM;//PCM采样
mAudioFormat.mFormatFlags = kAudioFormatFlagIsSignedInteger | kAudioFormatFlagIsPacked;
mAudioFormat.mFramesPerPacket = 1;//每个数据包多少帧
mAudioFormat.mChannelsPerFrame = 1;//1单声道，2立体声
mAudioFormat.mBitsPerChannel = 16;//语音每采样点占用位数
mAudioFormat.mBytesPerFrame = mAudioFormat.mBitsPerChannel*mAudioFormat.mChannelsPerFrame/8;//每帧的bytes数
mAudioFormat.mBytesPerPacket = mAudioFormat.mBytesPerFrame*mAudioFormat.mFramesPerPacket;//每个数据包的bytes总数，每帧的bytes数＊每个数据包的帧数
mAudioFormat.mReserved = 0;
CheckError(AudioUnitSetProperty(_toneUnit,
kAudioUnitProperty_StreamFormat,
kAudioUnitScope_Output, kInputBus,
&mAudioFormat, sizeof(mAudioFormat)),
"couldn't set the remote I/O unit's input client format");
CheckError(AudioUnitSetProperty(_toneUnit,
kAudioOutputUnitProperty_SetInputCallback,
kAudioUnitScope_Output,
kInputBus,
&_inputProc, sizeof(_inputProc)),
"couldnt set remote i/o render callback for input");
CheckError(AudioUnitInitialize(_toneUnit),
"couldn't initialize the remote I/O unit");
具体参数说明请参照iOS音频编程之变声处理
采集音频数据的输入回调
static OSStatus inputRenderTone(
void *inRefCon,
AudioUnitRenderActionFlags *ioActionFlags,
const AudioTimeStamp *inTimeStamp,
UInt32 inBusNumber,
UInt32 inNumberFrames,
AudioBufferList *ioData)
{
VoiceConvertHandle *THIS=(__bridge VoiceConvertHandle*)inRefCon;
AudioBufferList bufferList;
bufferList.mNumberBuffers = 1;
bufferList.mBuffers[0].mData = NULL;
bufferList.mBuffers[0].mDataByteSize = 0;
OSStatus status = AudioUnitRender(THIS->_toneUnit,
ioActionFlags,
inTimeStamp,
kInputBus,
inNumberFrames,
&bufferList);
NSInteger lastTimeRear = recordStruct.rear;
for (int i = 0; i < inNumberFrames; i++) {
SInt16 data = ((SInt16 *)bufferList.mBuffers[0].mData)[i];
recordStruct.recordArr[recordStruct.rear] = data;
recordStruct.rear = (recordStruct.rear+1)%kRecordDataLen;
}
if ((lastTimeRear/1024 + 1) == (recordStruct.rear/1024)) {
pthread_cond_signal(&recordCond);
}
return status;
}
采用循环队列存储原始的音频数据，每1024点的PCM数据，让Converter转换为AAC编码,所以当收集了1024点PCM后，唤醒Converter线程。
3)音频转码
初始化
AudioStreamBasicDescription sourceDes = mAudioFormat;
AudioStreamBasicDescription targetDes;
memset(&targetDes, 0, sizeof(targetDes));
targetDes.mFormatID = kAudioFormatMPEG4AAC;
targetDes.mSampleRate = kSmaple;
targetDes.mChannelsPerFrame = sourceDes.mChannelsPerFrame;
UInt32 size = sizeof(targetDes);
CheckError(AudioFormatGetProperty(kAudioFormatProperty_FormatInfo,
0, NULL, &size, &targetDes),
"couldnt create target data format");
//选择软件编码
AudioClassDescription audioClassDes;
CheckError(AudioFormatGetPropertyInfo(kAudioFormatProperty_Encoders,
sizeof(targetDes.mFormatID),
&targetDes.mFormatID,
&size), "cant get kAudioFormatProperty_Encoders");
UInt32 numEncoders = size/sizeof(AudioClassDescription);
AudioClassDescription audioClassArr[numEncoders];
CheckError(AudioFormatGetProperty(kAudioFormatProperty_Encoders,
sizeof(targetDes.mFormatID),
&targetDes.mFormatID,
&size,
audioClassArr),
"wrirte audioClassArr fail");
for (int i = 0; i < numEncoders; i++) {
if (audioClassArr[i].mSubType == kAudioFormatMPEG4AAC
&& audioClassArr[i].mManufacturer == kAppleSoftwareAudioCodecManufacturer) {
memcpy(&audioClassDes, &audioClassArr[i], sizeof(AudioClassDescription));
break;
}
}
CheckError(AudioConverterNewSpecific(&sourceDes, &targetDes, 1,
&audioClassDes, &_encodeConvertRef),
"cant new convertRef");
size = sizeof(sourceDes);
CheckError(AudioConverterGetProperty(_encodeConvertRef, kAudioConverterCurrentInputStreamDescription, &size, &sourceDes), "cant get kAudioConverterCurrentInputStreamDescription");
size = sizeof(targetDes);
CheckError(AudioConverterGetProperty(_encodeConvertRef, kAudioConverterCurrentOutputStreamDescription, &size, &targetDes), "cant get kAudioConverterCurrentOutputStreamDescription");
UInt32 bitRate = 64000;
size = sizeof(bitRate);
CheckError(AudioConverterSetProperty(_encodeConvertRef,
kAudioConverterEncodeBitRate,
size, &bitRate),
"cant set covert property bit rate");
[self performSelectorInBackground:@selector(convertPCMToAAC) withObject:nil];
主要是设置编码器的输入音频格式(PCM),输出音频格式(AAC),选择软件编码器(默认使用硬件编码器),设置编码器的比特率
AAC编码
-(void)convertPCMToAAC{
UInt32 maxPacketSize = 0;
UInt32 size = sizeof(maxPacketSize);
CheckError(AudioConverterGetProperty(_encodeConvertRef,
kAudioConverterPropertyMaximumOutputPacketSize,
&size,
&maxPacketSize),
"cant get max size of packet");
AudioBufferList *bufferList = malloc(sizeof(AudioBufferList));
bufferList->mNumberBuffers = 1;
bufferList->mBuffers[0].mNumberChannels = 1;
bufferList->mBuffers[0].mData = malloc(maxPacketSize);
bufferList->mBuffers[0].mDataByteSize = maxPacketSize;
for (; ; ) {
@autoreleasepool {
pthread_mutex_lock(&recordLock);
while (ABS(recordStruct.rear - recordStruct.front) < 1024 ) {
pthread_cond_wait(&recordCond, &recordLock);
}
pthread_mutex_unlock(&recordLock);
SInt16 *readyData = (SInt16 *)calloc(1024, sizeof(SInt16));
memcpy(readyData, &recordStruct.recordArr[recordStruct.front], 1024*sizeof(SInt16));
recordStruct.front = (recordStruct.front+1024)%kRecordDataLen;
UInt32 packetSize = 1;
AudioStreamPacketDescription *outputPacketDescriptions = malloc(sizeof(AudioStreamPacketDescription)*packetSize);
bufferList->mBuffers[0].mDataByteSize = maxPacketSize;
CheckError(AudioConverterFillComplexBuffer(_encodeConvertRef,
encodeConverterComplexInputDataProc,
readyData,
&packetSize,
bufferList,
outputPacketDescriptions),
"cant set AudioConverterFillComplexBuffer");
free(outputPacketDescriptions);
free(readyData);
NSMutableData *fullData = [NSMutableData dataWithBytes:bufferList->mBuffers[0].mData length:bufferList->mBuffers[0].mDataByteSize];
if ([self.delegate respondsToSelector:@selector(covertedData:)]) {
[self.delegate covertedData:[fullData copy]];
}
}
}
新建的bufferList是用来存放每次转码后的AAC音频数据.for循环中等待音频输入回调存满1024个PCM数组并唤醒它。outputPacketDescriptions数组是每次转换的AAC编码后各个包的描述,但这里每次只转换一包数据(由传入的packetSize决定)。调用AudioConverterFillComplexBuffer触发转码，他的第二个参数是填充原始音频数据的回调。转码完成后，会将转码的数据存放在它的第五个参数中(bufferList).转换完成的AAC就可以发送给另外一台手机了。
填充原始数据回调
OSStatus encodeConverterComplexInputDataProc(AudioConverterRef inAudioConverter,
UInt32 *ioNumberDataPackets,
AudioBufferList *ioData,
AudioStreamPacketDescription **outDataPacketDescription,
void *inUserData)
{
ioData->mBuffers[0].mData = inUserData;
ioData->mBuffers[0].mNumberChannels = 1;
ioData->mBuffers[0].mDataByteSize = 1024*2;
*ioNumberDataPackets = 1024;
return 0;
}
4）Audio Queue播放AAC音频数据
Audio Queue基础知识
音频数据以一个个AudioQueueBuffer的形式存在与音频队列中，Audio Queue使用它提供的音频数据来播放，某一个AudioQueueBuffer使用完毕后，会调用Audio Queue的回调，要求用户再在这个AudioQueueBuffer填入数据，并使它加入Audio Queue中，如此循环，达到不间断播放音频数据的效果。
Audio Queue初始化
CheckError(AudioQueueNewOutput(&targetDes,
fillBufCallback,
(__bridge void *)self,
NULL,
NULL,
0,
&(_playQueue)),
"cant new audio queue");
CheckError( AudioQueueSetParameter(_playQueue,
kAudioQueueParam_Volume, 1.0),
"cant set audio queue gain");
for (int i = 0; i < 3; i++) {
AudioQueueBufferRef buffer;
CheckError(AudioQueueAllocateBuffer(_playQueue, 1024, &buffer), "cant alloc buff");
BNRAudioQueueBuffer *buffObj = [[BNRAudioQueueBuffer alloc] init];
buffObj.buffer = buffer;
[_buffers addObject:buffObj];
[_reusableBuffers addObject:buffObj];
}
[self performSelectorInBackground:@selector(playData) withObject:nil];
Audio Queue播放音频数据
-(void)playData{
for (; ; ) {
@autoreleasepool {
NSMutableData *data = [[NSMutableData alloc] init];
pthread_mutex_lock(&playLock);
if (self.aacArry.count%8 != 0 || self.aacArry.count == 0) {
pthread_cond_wait(&playCond, &playLock);
}
AudioStreamPacketDescription *paks = calloc(sizeof(AudioStreamPacketDescription), 8);
for (int i = 0; i < 8 ; i++) {//8包AAC数据组成放入一个AudioQueueBuffer的数据包
BNRAudioData *audio = [self.aacArry firstObject];
[data appendData:audio.data];
paks[i].mStartOffset = audio.packetDescription.mStartOffset;
paks[i].mDataByteSize = audio.packetDescription.mDataByteSize;
[self.aacArry removeObjectAtIndex:0];
}
pthread_mutex_unlock(&playLock);
pthread_mutex_lock(&buffLock);
if (_reusableBuffers.count == 0) {
static dispatch_once_t onceToken;
dispatch_once(&onceToken, ^{
AudioQueueStart(_playQueue, nil);
});
pthread_cond_wait(&buffcond, &buffLock);
}
BNRAudioQueueBuffer *bufferObj = [_reusableBuffers firstObject];
[_reusableBuffers removeObject:bufferObj];
pthread_mutex_unlock(&buffLock);
memcpy(bufferObj.buffer->mAudioData,[data bytes] , [data length]);
bufferObj.buffer->mAudioDataByteSize = (UInt32)[data length];
CheckError(AudioQueueEnqueueBuffer(_playQueue, bufferObj.buffer, 8, paks), "cant enqueue");
free(paks);
}
}
}
static void fillBufCallback(void *inUserData,
AudioQueueRef inAQ,
AudioQueueBufferRef buffer){
VoiceConvertHandle *THIS=(__bridge VoiceConvertHandle*)inUserData;
for (int i = 0; i < THIS->_buffers.count; ++i) {
if (buffer == [THIS->_buffers[i] buffer]) {
pthread_mutex_lock(&buffLock);
[THIS->_reusableBuffers addObject:THIS->_buffers[i]];
pthread_mutex_unlock(&buffLock);
pthread_cond_signal(&buffcond);
break;
}
}
}
在playData中等待收到的aacArry数据，这里要注意:每1024点PCM转换成的一包AAC数据加入到AudioQueueBuffer中，不足以使Audio Queue播放音频，所以这里使用8包AAC数据放到一个AudioQueueBuffer中。fillBufCallback是Audio Queue播放完一个AudioQueueBuffer调用的回调函数，在这里面通知playData可以往使用完的AudioQueueBufferRef填数据了，填完后，用AudioQueueEnqueueBuffer将它加入Audio Queue中，这个三个AudioQueueBufferRef不断重用。
实时语音通信处理
原来是想用蓝牙来传送数据的，但是自己写的蓝牙传送数据机制的速度跟不上转换的AAC数据。使用
MultipeerConnectivity框架既可使用蓝牙也可以使用WIFI来通信，底层自动选择。当把两个手机的WIFI都关掉时，他们使用蓝牙来传送数据，在刚刚建立通话时，能听到传送的语音，之后就听不到了，使用wifi传输数据时不会出现这种情况。
- MultipeerConnectivity基础知识
MCNearbyServiceAdvertiser发送广播，并接收MCNearbyServiceBrowser端的邀请,MCSession发送接收数据、管理连接状态。建立连接和通信的流程是，MCNearbyServiceAdvertiser广播服务，MCNearbyServiceBrowser搜到这个服务后，要求把这个服务所对用的MCPeerID加入到它自己(MCNearbyServiceBrowser端)的MCSession中，MCNearbyServiceAdvertiser收到这个邀请，并同意，同时也将MCNearbyServiceBrowser端对应的MCPeerID加入到了它自己(MCNearbyServiceAdvertiser)的MCSession中.
之后双方可以使用各自的MCSession发送接收数据。
2）各端发送本身转码的AAC数据，并接收对方发送的AAC数据提供给Auduio queue播放
