如何使用 AVFoundation 以正确的音高播放不同采样率的音频文件？

Posted 2023-02-25

【中文标题】如何使用 AVFoundation 以正确的音高播放不同采样率的音频文件？

【英文标题】：How can audio files with different sampling rates be played at the correct pitch using AVFoundation?

【发布时间】：2021-07-06 23:02:48

【问题描述】：

AVAudioEngine 和 AVAudioPlayerNodes 在采样率不同时，应如何配置它们以正确的音高播放音频文件？

我读到混合器可以处理采样率转换，但我还没有实现这种行为。我正在使用扩展来播放循环音频片段，播放器应该可以使用压缩和 PCM 文件；我不知道这是否会决定解决方案。

我尝试在 AVAudioPlayerNode 的 installTap(onBus:bufferSize:format: block:) 块内使用 AVAudioConverter，但遇到了各种崩溃，并且不确定这是否是正确的解决方案。我是在正确的轨道上还是有更简单的解决方案？

import AVFoundation
import SwiftUI

@main
struct SampleRateMixerApp: App 

    private let audioEngine = AVAudioEngine()
    private let playerA = AVAudioPlayerNode()
    private let playerB = AVAudioPlayerNode()

    var body: some Scene 
        WindowGroup 
            Button("Play")  try? playFiles() 
            Button("Stop")  playerA.stop(); playerB.stop() 
        
    

    func playFiles() throws 
        _ = audioEngine.mainMixerNode
        try audioEngine.start()
        audioEngine.prepare()

        let pathA = Bundle.main.path(forResource: "32khz_sample_rate", ofType: "aif")!
        try setupPlayerNode(player: playerA, withAudioEngine: audioEngine, atPath: pathA)

        let pathB = Bundle.main.path(forResource: "48khz_sample_rate", ofType: "mp3")!
        try setupPlayerNode(player: playerB, withAudioEngine: audioEngine, atPath: pathB)
    

    func setupPlayerNode(player: AVAudioPlayerNode, withAudioEngine engine: AVAudioEngine, atPath path: String) throws 
        engine.attach(player)
        engine.connect(player, to: engine.mainMixerNode, format: nil)

        let url = URL(string: path)!
        let file = try AVAudioFile(forReading: url)
        let frameCount = AVAudioFrameCount(file.length)
        let buffer = AVAudioPCMBuffer(pcmFormat: file.processingFormat, frameCapacity: frameCount)!
        try file.read(into: buffer)

        player.scheduleBufferSegment(buffer, range: 0.25...0.75, looping: true)
        player.play()
    


extension AVAudioPlayerNode 

    func scheduleBufferSegment(_ buffer: AVAudioPCMBuffer, range: ClosedRange<Double>, looping: Bool) 
        let length = Double(buffer.frameLength)
        let startFrame = AVAudioFramePosition(range.lowerBound * length)
        let endFrame = AVAudioFramePosition(range.upperBound * length)

        guard let bufferSegment = buffer.segment(from: startFrame, to: endFrame) else  return 

        if looping 
            scheduleBuffer(bufferSegment, at: nil, options: [.loops, .interrupts])
         else 
            scheduleBuffer(bufferSegment, at: nil, options: [.interrupts]) 
                DispatchQueue.main.async  [weak self] in
                    self?.stop()
                
            
        
    


extension AVAudioPCMBuffer 

    func segment(from startFrame: AVAudioFramePosition, to endFrame: AVAudioFramePosition) -> AVAudioPCMBuffer? 
        let framesToCopy = AVAudioFrameCount(endFrame - startFrame)
        guard let segment = AVAudioPCMBuffer(pcmFormat: format, frameCapacity: framesToCopy) else  return nil 

        let sourcePointer = UnsafeMutableAudioBufferListPointer(mutableAudioBufferList)
        let destinationPointer = UnsafeMutableAudioBufferListPointer(segment.mutableAudioBufferList)
        let sampleSize = format.streamDescription.pointee.mBytesPerFrame

        for (source, destination) in zip(sourcePointer, destinationPointer) 
            memcpy(destination.mData,
                   source.mData?.advanced(by: Int(startFrame) * Int(sampleSize)),
                   Int(framesToCopy) * Int(sampleSize))
        

        segment.frameLength = framesToCopy
        return segment
    

【参考方案1】：

在加载每个文件后调用AVAudioEngine 的connect(:to:format:) 传入AVAudioFiles 的processingFormat。这可确保AVAudioPlayerNode 的输出设置为与其将播放的文件相同的采样率。

以下代码说明了这是如何完成的。该问题与播放缓冲区段无关，因此未将其包含在解决方案中。虽然音频可以在没有先断开播放器的情况下正常播放，但我怀疑它可能会泄漏内存，所以为了安全起见先断开节点。

import AVFoundation

let audioEngine = AVAudioEngine()
let player = AVAudioPlayerNode()

func playFiles() throws 
    _ = audioEngine.mainMixerNode // Called to ensure mainMixerNode singleton is instantiated
    try audioEngine.start()
    audioEngine.prepare()
    
    audioEngine.attach(player)
    
    let path = Bundle.main.path(forResource: "32khz", ofType: "aif")!
    let url = URL(string: path)!
    let file = try AVAudioFile(forReading: url)
    let frameCount = AVAudioFrameCount(file.length)
    let buffer = AVAudioPCMBuffer(pcmFormat: file.processingFormat, frameCapacity: frameCount)!
    try file.read(into: buffer)
    
    audioEngine.disconnectNodeOutput(player) // May be unnecessary 
    audioEngine.connect(player, to: audioEngine.mainMixerNode, format: file.processingFormat)
    player.scheduleBuffer(buffer)
    player.play()