。
我正在开发一个iOS应用程序，它由2个主模块组成：核心音频上的音频分析模块基础，以及使用AudioKit的输出模块。
这是音频输入类：
import AVFoundation
typealias AudioInputCallback = (
_ timeStamp: Double,
_ numberOfFrames: Int,
_ samples: [Float]
) -> Void
/// Sets up an audio input session and notifies when new buffer data is available.
class AudioInputUtility: NSObject {
private(set) var audioUnit: AudioUnit!
var audioSession : AVAudioSession = AVAudioSession.sharedInstance()
var sampleRate: Float
var numberOfChannels: Int
/// When true, performs DC offset rejection on the incoming buffer before invoking the audioInputCallback.
var shouldPerformDCOffsetRejection: Bool = false
private let outputBus: UInt32 = 0
private let inputBus: UInt32 = 1
private var audioInputCallback: AudioInputCallback!
/// Instantiate a AudioInput.
/// - Parameter audioInputCallback: Invoked when audio data is available.
/// - Parameter sampleRate: The sample rate to set up the audio session with.
/// - Parameter numberOfChannels: The number of channels to set up the audio session with.
init(audioInputCallback callback: @escaping AudioInputCallback, sampleRate: Float = 44100.0, numberOfChannels: Int = 1) { // default values if not specified
self.sampleRate = sampleRate
self.numberOfChannels = numberOfChannels
audioInputCallback = callback
}
/// Start recording. Prompts for access to microphone if necessary.
func startRecording() {
do {
if self.audioUnit == nil {
setupAudioSession()
setupAudioUnit()
}
try self.audioSession.setActive(true)
var osErr: OSStatus = 0
osErr = AudioUnitInitialize(self.audioUnit)
assert(osErr == noErr, "*** AudioUnitInitialize err (osErr)")
osErr = AudioOutputUnitStart(self.audioUnit)
assert(osErr == noErr, "*** AudioOutputUnitStart err (osErr)")
} catch {
print("*** startRecording error: (error)")
}
}
/// Stop recording.
func stopRecording() {
do {
var osErr: OSStatus = 0
osErr = AudioOutputUnitStop(self.audioUnit)
osErr = AudioUnitUninitialize(self.audioUnit)
assert(osErr == noErr, "*** AudioUnitUninitialize err (osErr)")
try self.audioSession.setActive(false)
} catch {
print("*** error: (error)")
}
}
private let recordingCallback: AURenderCallback = { (inRefCon, ioActionFlags, inTimeStamp, inBusNumber, inNumberFrames, ioData) -> OSStatus in
let audioInput = unsafeBitCast(inRefCon, to: AudioInputUtility.self)
var osErr: OSStatus = 0
// We've asked CoreAudio to allocate buffers for us, so just set mData to nil and it will be populated on AudioUnitRender().
var bufferList = AudioBufferList(
mNumberBuffers: 1,
mBuffers: AudioBuffer(
mNumberChannels: UInt32(audioInput.numberOfChannels),
mDataByteSize: 4,
mData: nil))
osErr = AudioUnitRender(audioInput.audioUnit,
ioActionFlags,
inTimeStamp,
inBusNumber,
inNumberFrames,
&bufferList)
assert(osErr == noErr, "*** AudioUnitRender err (osErr)")
// Move samples from mData into our native [Float] format.
var monoSamples = [Float]()
let ptr = bufferList.mBuffers.mData?.assumingMemoryBound(to: Float.self)
monoSamples.append(contentsOf: UnsafeBufferPointer(start: ptr, count: Int(inNumberFrames)))
if audioInput.shouldPerformDCOffsetRejection {
DCRejectionFilterProcessInPlace(&monoSamples, count: Int(inNumberFrames))
}
// Not compatible with Obj-C...
audioInput.audioInputCallback(inTimeStamp.pointee.mSampleTime / Double(audioInput.sampleRate),
Int(inNumberFrames),
monoSamples)
return 0
}
private func setupAudioSession() {
if !audioSession.availableCategories.contains(AVAudioSessionCategoryRecord) {
print("can't record! bailing.")
return
}
do {
//https://developer.apple.com/reference/avfoundation/avaudiosession/1669963-audio_session_categories
try audioSession.setCategory(AVAudioSessionCategoryRecord)
// "Appropriate for applications that wish to minimize the effect of system-supplied signal processing for input and/or output audio signals."
// NB: This turns off the high-pass filter that CoreAudio normally applies.
try audioSession.setMode(AVAudioSessionModeMeasurement)
try audioSession.setPreferredSampleRate(Double(sampleRate))
// NB: This is considered a 'hint' and more often than not is just ignored.
// number of seconds to record -> voglio 1024 samples
try audioSession.setPreferredIOBufferDuration(0.05)
audioSession.requestRecordPermission { (granted) -> Void in
if !granted {
print("*** record permission denied")
}
}
} catch {
print("*** audioSession error: (error)")
}
}
private func setupAudioUnit() {
var componentDesc:AudioComponentDescription = AudioComponentDescription(
componentType: OSType(kAudioUnitType_Output),
componentSubType: OSType(kAudioUnitSubType_RemoteIO), // Always this for iOS.
componentManufacturer: OSType(kAudioUnitManufacturer_Apple),
componentFlags: 0,
componentFlagsMask: 0)
var osErr: OSStatus = 0
// Get an audio component matching our description.
let component: AudioComponent! = AudioComponentFindNext(nil, &componentDesc)
assert(component != nil, "Couldn't find a default component")
// Create an instance of the AudioUnit
var tempAudioUnit: AudioUnit?
osErr = AudioComponentInstanceNew(component, &tempAudioUnit)
self.audioUnit = tempAudioUnit
assert(osErr == noErr, "*** AudioComponentInstanceNew err (osErr)")
// Enable I/O for input.
var one:UInt32 = 1
osErr = AudioUnitSetProperty(audioUnit,
kAudioOutputUnitProperty_EnableIO,
kAudioUnitScope_Input,
inputBus,
&one,
UInt32(MemoryLayout<UInt32>.size))
assert(osErr == noErr, "*** AudioUnitSetProperty err (osErr)")
osErr = AudioUnitSetProperty(audioUnit,
kAudioOutputUnitProperty_EnableIO,
kAudioUnitScope_Output,
outputBus,
&one,
UInt32(MemoryLayout<UInt32>.size))
assert(osErr == noErr, "*** AudioUnitSetProperty err (osErr)")
// Set format to 32 bit, floating point, linear PCM
var streamFormatDesc:AudioStreamBasicDescription = AudioStreamBasicDescription(
mSampleRate: Double(sampleRate),
mFormatID: kAudioFormatLinearPCM,
mFormatFlags: kAudioFormatFlagsNativeFloatPacked | kAudioFormatFlagIsNonInterleaved, // floating point data - docs say this is fastest
mBytesPerPacket: 4,
mFramesPerPacket: 1,
mBytesPerFrame: 4,
mChannelsPerFrame: UInt32(self.numberOfChannels),
mBitsPerChannel: 4 * 8,
mReserved: 0
)
// Set format for input and output busses
osErr = AudioUnitSetProperty(audioUnit,
kAudioUnitProperty_StreamFormat,
kAudioUnitScope_Input, outputBus,
&streamFormatDesc,
UInt32(MemoryLayout<AudioStreamBasicDescription>.size))
assert(osErr == noErr, "*** AudioUnitSetProperty err (osErr)")
osErr = AudioUnitSetProperty(audioUnit,
kAudioUnitProperty_StreamFormat,
kAudioUnitScope_Output,
inputBus,
&streamFormatDesc,
UInt32(MemoryLayout<AudioStreamBasicDescription>.size))
assert(osErr == noErr, "*** AudioUnitSetProperty err (osErr)")
// Set up our callback.
var inputCallbackStruct = AURenderCallbackStruct(inputProc: recordingCallback, inputProcRefCon: UnsafeMutableRawPointer(Unmanaged.passUnretained(self).toOpaque()))
osErr = AudioUnitSetProperty(audioUnit,
AudioUnitPropertyID(kAudioOutputUnitProperty_SetInputCallback),
AudioUnitScope(kAudioUnitScope_Global),
inputBus,
&inputCallbackStruct,
UInt32(MemoryLayout<AURenderCallbackStruct>.size))
assert(osErr == noErr, "*** AudioUnitSetProperty err (osErr)")
// Ask CoreAudio to allocate buffers for us on render. (This is true by default but just to be explicit about it...)
osErr = AudioUnitSetProperty(audioUnit,
AudioUnitPropertyID(kAudioUnitProperty_ShouldAllocateBuffer),
AudioUnitScope(kAudioUnitScope_Output),
inputBus,
&one,
UInt32(MemoryLayout<UInt32>.size))
assert(osErr == noErr, "*** AudioUnitSetProperty err (osErr)")
}
}
private func DCRejectionFilterProcessInPlace(_ audioData: inout [Float], count: Int) {
let defaultPoleDist: Float = 0.975
var mX1: Float = 0
var mY1: Float = 0
for i in 0..<count {
let xCurr: Float = audioData[i]
audioData[i] = audioData[i] - mX1 + (defaultPoleDist * mY1)
mX1 = xCurr
mY1 = audioData[i]
}
}
这是输出类：
private func initPlayer(){
do{
/*
let audioSession : AVAudioSession = AVAudioSession.sharedInstance()
//try audioSession.setActive(false)
try audioSession.setCategory(AVAudioSessionCategoryPlayback)
*/
// http://audiokit.io/playgrounds/Playback/Reading%20and%20Writing%20Audio%20Files/
let file = try AKAudioFile(readFileName: self.soundPath,
baseDir: .resources)
self.player = try AKAudioPlayer(file: file)
//player options
self.player!.looping = true
AKSettings.playbackWhileMuted = true
try AKSettings.setSession(category: .playback)
AudioKit.output = self.player
}catch{
print("Unresolved error (error)")
}
}
public func stopMaskingSound(){
if(player!.isPlaying){
self.player!.stop()
}
if audioKitIsStarted == true{
AudioKit.stop()
self.audioKitIsStarted = false
}
}
您可以看到音频输入和输出由2个不同的类管理。
我遇到的问题是，如果我执行此步骤：1）初始玩家和记录 ->停止它2）播放输出 ->停止3）Reinit Player
在第三步，我有以下例外：
[central] 54: ERROR: [0x16dfc3000] >avae> AVAudioIONodeImpl.mm:365: _GetHWFormat: required condition is false: hwFormat
*** Terminating app due to uncaught exception 'com.apple.coreaudio.avfaudio', reason: 'required condition is false: hwFormat'
有人知道它与什么相关吗？Audiokit＆lt; -> Core Audio是否有生命周期问题？
停止并重新启动音频单元可能会出现问题，因为音频过程的某些部分确实在另一个线程或线程中停止。周围的一项可能的工作可能是在停止和重新启动之间延迟约1秒钟，以便在尝试重新启动之前，将远程io异步滑到停止。