SwiftUI+AudioKit iOS应用从后台打开时卡顿问题求助

算力阿峰

2026-6-22

iOS端AudioKit调音器后台恢复卡顿问题排查求助

我正在开发一款支持macOS和iOS的调音器应用，使用AudioKit处理音频逻辑。macOS端运行正常，但iOS端将应用置于后台后再切回前台时，约有半数概率出现卡顿，后台停留时间越长触发概率越高。

推测卡顿原因是：AudioKit的AudioEngine从后台恢复后未正常启动，导致应用无法获取音频数据，进而停止UI更新。我已尝试三种音频引擎启停方案，均未解决问题：

使用.onAppear/.onDisappear控制引擎启停
监听scenePhase状态切换控制引擎
保持引擎持续运行

可复现问题的完整代码

import AudioKit
import SoundpipeAudioKit // PitchTap
import AudioKitEX // Fader
import AVFoundation // AVAudioSession
import SwiftUI

struct ContentView: View {
    @StateObject var td = ToneDetector()
    var body: some View {
        VStack {
            Text(td.data.note)
                .font(.system(size: 100, design: .serif))
        }
        .task {
            await PermissionsChecker.getMicrophoneAccess()
        }
        .task {
            if !td.engine.avEngine.isRunning {
                td.start()
            }
        }
    }
}

struct TunerData {
    var note = "-"
}

class ToneDetector : ObservableObject, HasAudioEngine {
    @Published var data = TunerData()
    let engine = AudioEngine()
    
    let mic: AudioEngine.InputNode
    let tappableA: Fader
    let tappableB: Fader
    let silence: Fader
    var tracker: PitchTap!
    
    let noteFrequencies: [Float] = [16.35, 17.32, 18.35, 19.45, 20.6, 21.83, 23.12, 24.5, 25.96, 27.5, 29.14, 30.87]
    let noteNamesFlats = ["C", "C♯", "D", "D♯", "E", "F", "F♯", "G", "G♯", "A", "A♯", "B"]
    let noteNamesSharps = ["C", "D♭", "D", "E♭", "E", "F", "G♭", "G", "A♭", "A", "B♭", "B"]
    
    init() {
        guard let inputTmp = engine.input else { fatalError() }
        mic = inputTmp

        tappableA = Fader(mic)
        tappableB = Fader(tappableA)
        silence = Fader(tappableB, gain: 0)
        engine.output = silence
        
        PermissionsChecker.setSessionParameters()
        
        tracker = PitchTap(mic) { pitch, amp in
            DispatchQueue.main.async {
                self.update(pitch[0], amp[0])
            }
        }
        tracker.start()
    }
    
    func update(_ pitch: AUValue, _ amp: AUValue) {
        guard amp > 0.07 else { return }
        
        // Get the frequency down or up to the range that we know
        var frequency = pitch
        while frequency > Float(noteFrequencies[noteFrequencies.count - 1]) {
            frequency /= 2.0
        }
        while frequency < Float(noteFrequencies[0]) {
            frequency *= 2.0
        }
        
        // Find the known note frequency we are CLOSEST to
        var minDistance: Float = 10000.0
        var index = 0
        for possibleIndex in 0..<noteFrequencies.count {
            let distance = fabsf(Float(noteFrequencies[possibleIndex]) - frequency)
            if distance < minDistance {
                index = possibleIndex
                minDistance = distance
            }
        }
        if noteNamesFlats[index] == noteNamesSharps[index] {
            data.note = "\(noteNamesFlats[index])"
        } else {
            data.note = "\(noteNamesFlats[index]) / \(noteNamesSharps[index])"
        }
    }
}

class PermissionsChecker {
    
    static func getMicrophoneAccess() async {
        if #available(iOS 17.0, *) {
            let permission = AVAudioApplication.shared.recordPermission
            switch permission {
                case .granted: return
                case .denied: print("Microphone permission not granted.")
                case .undetermined: break
                default: break
            }
            
            await AVAudioApplication.requestRecordPermission()
        }
    }
    
    static func setSessionParameters() {
        #if os(iOS)
            do {
                Settings.bufferLength = .short
                try AVAudioSession.sharedInstance().setPreferredIOBufferDuration(Settings.bufferLength.duration)
                try AVAudioSession.sharedInstance().setCategory(.playAndRecord, mode: .default, options: [.defaultToSpeaker, .mixWithOthers, .allowBluetooth])
                try AVAudioSession.sharedInstance().setActive(true)
            } catch {
                print("Failed to configure AVAudioSession.")
            }
        #endif
    }
}

已尝试的三种方案代码

尝试1：使用.onAppear/.onDisappear

struct ContentView: View {
    let engine = AudioEngine()
    var body: some View {
        VStack {
            ...
        }
        .onAppear() {
            engine.start()
        }
        .onDisappear() {
            engine.stop()
        }
    }
}

尝试2：监听scenePhase

struct ContentView: View {
    let engine = AudioEngine()
    @Environment(\.scenePhase) var scenePhase
    var body: some View {
        VStack {
            ...
        }
        .onChange(of: scenePhase) { oldPhase, newPhase in
            if newPhase == .active {
                if !engine.avEngine.isRunning {
                    engine.start()
                }
            } else if newPhase == .inactive {
                if engine.avEngine.isRunning {
                    engine.stop()
                }
            } else if newPhase == .background {
                if engine.avEngine.isRunning {
                    engine.stop()
                }
            }
        }
    }
}

尝试3：保持引擎持续运行

struct ContentView: View {
    let engine = AudioEngine()
    var body: some View {
        VStack {
            ...
        }
        .task {
            if !engine.avEngine.isRunning {
                engine.start()
            }
        }
    }
}

排查方向与解决思路

同步AVAudioSession与AudioEngine状态
iOS后台时AVAudioSession可能被系统暂停，恢复时不要依赖avEngine.isRunning的判断（该值可能与实际session状态不一致）。需先重新激活session，再启动引擎，并处理激活失败的情况：

func restartAudioEngine() {
    #if os(iOS)
        do {
            try AVAudioSession.sharedInstance().setActive(true, options: .notifyOthersOnDeactivation)
            if !engine.isRunning {
                try engine.start()
                tracker?.start() // 重启PitchTap监听
            }
        } catch {
            print("重启音频引擎失败：\(error)")
        }
    #endif
}

重启PitchTap监听
后台状态下PitchTap可能被终止，仅重启引擎无法恢复音频数据监听。需在应用恢复前台时，手动重启PitchTap。
调整引擎启停策略
避免在background/inactive状态下手动停止引擎，让系统自动暂停音频会话；恢复前台时，通过系统通知触发音频链路的重新初始化，减少手动启停带来的状态不一致问题。

优化异步启动时序
当前的.task启动逻辑可能与权限请求、session设置存在时序冲突。需确保麦克风权限获取完成后，再启动引擎：

.task {
    await PermissionsChecker.getMicrophoneAccess()
    if !td.engine.isRunning {
        do {
            try td.start()
        } catch {
            print("启动引擎失败：\(error)")
        }
    }
}

监听音频会话通知
注册AVAudioSession.interruptionNotification和AVAudioSession.routeChangeNotification，在收到中断结束通知时重新初始化音频组件：

// 在ToneDetector的init中添加监听
NotificationCenter.default.addObserver(self, selector: #selector(handleAudioInterruption(_:)), name: AVAudioSession.interruptionNotification, object: nil)

@objc func handleAudioInterruption(_ notification: Notification) {
    guard let info = notification.userInfo,
          let typeValue = info[AVAudioSessionInterruptionTypeKey] as? UInt,
          let type = AVAudioSession.InterruptionType(rawValue: typeValue) else { return }
    
    if type == .ended {
        restartAudioEngine()
    }
}

检查AudioKit版本
确认当前使用的AudioKit版本是否存在后台恢复的已知bug，尝试升级到最新稳定版。

内容的提问来源于stack exchange，提问作者Sophie_S

SwiftUI+AudioKit iOS应用从后台打开时卡顿问题求助

可复现问题的完整代码 #

已尝试的三种方案代码 #

尝试1：使用.onAppear/.onDisappear #

尝试2：监听scenePhase #

尝试3：保持引擎持续运行 #