iOS平台语音边录边压边上传的实现

本文只讲述iOS平台如何实现边录制音频边压缩编码边上传到服务端，至于播放则只下载解码播放并不涉及

录制

使用AudioToolbox.framework -> AudioQueue类进行录音/播放



三个缓冲器Buffers:每个缓冲器都是一个存储音频数据的临时仓库。
一个缓冲队列Buffer Queue:一个包含音频缓冲器的有序队列。
一个回调Callback:一个自定义的队列回调函数。

- (void)_createAudioInputQueue
{
APMM_DEBUG(@"WAudioInputQueue _createAudioInputQueue");
if (![self _checkAudioQueueSuccess:AudioQueueNewInput(&_format, MCAudioQueueInuputCallback, (__bridge void *)(self), CFRunLoopGetMain(), NULL, 0, &_audioQueue)])
{
return;
}
AudioQueueAddPropertyListener(_audioQueue, kAudioQueueProperty_IsRunning, WAudioInputQueuePropertyCallback, (__bridge void *)(self));
_meterState = (AudioQueueLevelMeterState *)calloc(sizeof(AudioQueueLevelMeterState),_format.mChannelsPerFrame);
UInt32 trueValue = true;
AudioQueueSetProperty(_audioQueue,kAudioQueueProperty_EnableLevelMetering, &trueValue, sizeof(UInt32));

for (int i = 0; i < MCAudioQueueBufferCount; ++i)
{
AudioQueueBufferRef buffer;
if (![self _checkAudioQueueSuccess:AudioQueueAllocateBuffer(_audioQueue, _bufferSize, &buffer)])
{
break;
}

if (![self _checkAudioQueueSuccess:AudioQueueEnqueueBuffer(_audioQueue, buffer, 0, NULL)])
{
break;
}
}
}

static void MCAudioQueueInuputCallback(void *inClientData,
AudioQueueRef inAQ,
AudioQueueBufferRef inBuffer,
const AudioTimeStamp *inStartTime,
UInt32 inNumberPacketDescriptions,
const AudioStreamPacketDescription *inPacketDescs)
{
WAudioInputQueue *audioOutputQueue = (__bridge WAudioInputQueue *)inClientData;
[audioOutputQueue handleAudioQueueOutputCallBack:inAQ
buffer:inBuffer
inStartTime:inStartTime
inNumberPacketDescriptions:inNumberPacketDescriptions
inPacketDescs:inPacketDescs];
}

- (void)handleAudioQueueOutputCallBack:(AudioQueueRef)audioQueue
buffer:(AudioQueueBufferRef)buffer
inStartTime:(const AudioTimeStamp *)inStartTime
inNumberPacketDescriptions:(UInt32)inNumberPacketDescriptions
inPacketDescs:(const AudioStreamPacketDescription *)inPacketDescs
{
if (_started)
{
[_buffer appendBytes:buffer->mAudioData length:buffer->mAudioDataByteSize];
if ([_buffer length] >= _bufferSize)
{
NSRange range = NSMakeRange(0, _bufferSize);
NSData *subData = [_buffer subdataWithRange:range];
[_delegate inputQueue:self inputData:subData numberOfPackets:inNumberPacketDescriptions finish:NO];
[_buffer replaceBytesInRange:range withBytes:NULL length:0];
}
[self _checkAudioQueueSuccess:AudioQueueEnqueueBuffer(_audioQueue, buffer, 0, NULL)];
}else{
[_buffer appendBytes:buffer->mAudioData length:buffer->mAudioDataByteSize];
NSRange range = NSMakeRange(0, buffer->mAudioDataByteSize);
NSData *subData = [_buffer subdataWithRange:range];
[_delegate inputQueue:self inputData:subData numberOfPackets:inNumberPacketDescriptions finish:NO];
[_buffer replaceBytesInRange:range withBytes:NULL length:0];
}
APMM_DEBUG(@"handleAudioQueueOutputCallBack, data length:%u",(unsigned int)buffer->mAudioDataByteSize);
}

通过AudioQueue类的注册的callback方法拿到语音buffer（PCM 未压缩音频数据）

压缩

目前语音压缩的编码格式很多，像AMR、SILK等，但是在同等采样率和比特率的条件下，SILK在音质包括降噪等方面脚AMR较优秀，可以看到微信也是采用后者。



silk相关参数的配置

/* Define decode codec specific settings should be moved to h file */
#define DECODE_MAX_BYTES_PER_FRAME     1024
#define DECODE_MAX_INPUT_FRAMES        5
#define DECODE_MAX_FRAME_LENGTH        480
#define DECODE_FRAME_LENGTH_MS         20
#define DECODE_MAX_API_FS_KHZ          48
#define DECODE_MAX_LBRR_DELAY          2

/* Define encode codec specific settings */
#define ENCODE_MAX_BYTES_PER_FRAME     250 // Equals peak bitrate of 100 kbps
#define ENCODE_MAX_INPUT_FRAMES        5
#define ENCODE_FRAME_LENGTH_MS         20
#define ENCODE_MAX_API_FS_KHZ          48

在录制中拿到的语音buffer 通过SILK的如上配置进行编码最终拿到编码后的silk数据buffer,到此，我们已经能实时拿到录音的编码后数据。

上传
如果按照其他类型文件上传，语音需要拿最终结束录音后保存的文件进行上传，每次用户录音完成，如果录音时间较久，比如60秒后才进行上传，那么意味着60秒后才去建立连接上传服务端，那么其在数据网络或者弱网条件下的耗时是很大的，于是我们可以在用户点击录音开始便可以建连、每次通过AudioQueue拿到编码后的数据进行写数据，直到用户结束录音，我们的录音数据也同时上传完毕。



这里我们可以使用

HTTP/1.1 POST
multipart/form-data
首先，${bound} 是一个占位符，代表我们规定的分割符，可以自己任意规定，但为了避免和正常文本重复了，尽量要使用复杂一点的内容。
然后，Content-Type里指明了数据是以mutipart/form-data来编码，本次请求的boundary是什么内容。消息主体里按照字段个数又分为多个结构类似的部分，每部分都是以–boundary开始，紧接着内容描述信息，然后是回车，最后是字段具体内容（文本或二进制）。如果传输的是文件，还要包含文件名和文件类型信息。消息主体最后以–boundary–标示结束。

POST http://www.example.com HTTP/1.1
Content-Type:multipart/form-data; boundary=${bound}

--${bound}
Content-Disposition: form-data; name="text"

title

--${bound}
Content-Disposition: form-data; name="file"; filename="chrome.png"
Content-Type: image/png

PNG ... content of chrome.png ...
--${bound}--

在iOS平台， 已经有很多优秀的网络开源库，这里我采用的是AFNetWorking的第三方开源库，它本身就已经支持了multipart/form-data的协议封装，我们只需稍加改造，便可以通过HOLD住上传流来进行多表单方式的上传，一边录制一边压缩一边上传语音，直到结束录音没有语音数据。

- (NSMutableURLRequest *)multipartFormStreamRequestWithMethod:(NSString *)method
URLString:(NSString *)URLString
parameters:(NSDictionary *)parameters
constructingBodyWithBlock:(void (^)(id <AFMultipartFormData> formData))block
error:(NSError *__autoreleasing *)error
{
NSParameterAssert(method);
NSParameterAssert(![method isEqualToString:@"GET"] && ![method isEqualToString:@"HEAD"]);

NSMutableURLRequest *mutableRequest = [self requestWithMethod:method URLString:URLString parameters:nil error:error];

__block AFStreamingMultipartFormData *formData = [[AFStreamingMultipartFormData alloc] initWithURLRequest:mutableRequest stringEncoding:NSUTF8StringEncoding];

if (parameters) {
for (AFQueryStringPair *pair in AFQueryStringPairsFromDictionary(parameters)) {
NSData *data = nil;
if ([pair.value isKindOfClass:[NSData class]]) {
data = pair.value;
} else if ([pair.value isEqual:[NSNull null]]) {
data = [NSData data];
} else {
data = [[pair.value description] dataUsingEncoding:self.stringEncoding];
}

if (data) {
[formData appendPartWithFormData:data name:[pair.field description]];
}
}
}

if (block) {
block(formData);
}

return [formData requestByFinalizingMultipartFormDataWithOutLength];
}

- (NSMutableURLRequest *)requestByFinalizingMultipartFormDataWithOutLength {
if ([self.bodyStream isEmpty]) {
return self.request;
}

// Reset the initial and final boundaries to ensure correct Content-Length
[self.bodyStream setInitialAndFinalBoundaries];
[self.request setHTTPBodyStream:self.bodyStream];

[self.request setValue:[NSString stringWithFormat:@"multipart/form-data; boundary=%@", self.boundary] forHTTPHeaderField:@"Content-Type"];
[self.request setValue:@"100-Continue" forHTTPHeaderField:@"Expect"];
return self.request;
}

- (NSInteger)read:(uint8_t *)buffer
maxLength:(NSUInteger)length
{
if ([self streamStatus] == NSStreamStatusClosed) {
return 0;
}

NSInteger totalNumberOfBytesRead = 0;

#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wgnu"
while ((NSUInteger)totalNumberOfBytesRead < MIN(length, self.numberOfBytesInPacket)) {
if (!self.currentHTTPBodyPart || ![self.currentHTTPBodyPart hasBytesAvailable]) {
if (!(self.currentHTTPBodyPart = [self.HTTPBodyPartEnumerator nextObject])) {
break;
}
} else {
NSUInteger maxLength = length - (NSUInteger)totalNumberOfBytesRead;
NSInteger numberOfBytesRead = [self.currentHTTPBodyPart read:&buffer[totalNumberOfBytesRead] maxLength:maxLength];
if (numberOfBytesRead == -1) {
self.streamError = self.currentHTTPBodyPart.inputStream.streamError;
break;
}else if (numberOfBytesRead == 0){
[NSThread sleepForTimeInterval:1];
}else {
totalNumberOfBytesRead += numberOfBytesRead;

if (self.delay > 0.0f) {
[NSThread sleepForTimeInterval:self.delay];
}
return totalNumberOfBytesRead;
}
}
}
#pragma clang diagnostic pop
return totalNumberOfBytesRead;
}