如何使用AVAudioConverter从单声道转换为立体声？

Question

我正在尝试使用AVAudioEngine而不是AVAudioPlayer，因为在播放音频时，我需要对每个包进行一些处理，但在达到这个目的之前，我需要将16位的8khz单音频数据转换成立体声，这样AVAudioEngine就可以播放它了。这是我(不完整)做这件事的尝试。目前，我一直在研究如何使AVAudioConverter进行单立体声转换。如果我不使用AVAudioConverter，iOS运行时会抱怨输入格式与输出格式不匹配。如果我确实使用它(如下所示)，运行时不会抱怨，但是音频不能正常播放(可能是因为我没有正确地执行单立体声转换)。任何帮助都是非常感谢的！

  private func loadAudioData(audioData: Data?) {
      // Load audio data into player

      guard let audio = audioData else {return}
      do {
          let inputAudioFormat = AVAudioFormat(commonFormat: .pcmFormatInt16, sampleRate: Double(sampleRate), channels: 1, interleaved: false)
          let outputAudioFormat = self.audioEngine.mainMixerNode.outputFormat(forBus: 0)
          
          if inputAudioFormat != nil {
              let inputStreamDescription = inputAudioFormat?.streamDescription.pointee
              let outputStreamDescription = outputAudioFormat.streamDescription.pointee
              let count = UInt32(audio.count)
              if inputStreamDescription != nil && count > 0 {
                  if let ibpf = inputStreamDescription?.mBytesPerFrame {
                      let inputFrameCapacity = count / ibpf
                      let outputFrameCapacity = count / outputStreamDescription.mBytesPerFrame
                      self.pcmInputBuffer = AVAudioPCMBuffer(pcmFormat: inputAudioFormat!, frameCapacity: inputFrameCapacity)
                      self.pcmOutputBuffer = AVAudioPCMBuffer(pcmFormat: outputAudioFormat, frameCapacity: outputFrameCapacity)
          
                      if let input = self.pcmInputBuffer, let output = self.pcmOutputBuffer {
                          self.pcmConverter = AVAudioConverter(from: inputAudioFormat!, to: outputAudioFormat)
                          input.frameLength = input.frameCapacity
                      
                          let b = UnsafeMutableBufferPointer(start: input.int16ChannelData?[0], count: input.stride * Int(inputFrameCapacity))
                          let bytesCopied = audio.copyBytes(to: b)
                          assert(bytesCopied == count)
          
                          audioEngine.attach(playerNode)
                          audioEngine.connect(playerNode, to: audioEngine.mainMixerNode, format: nil)
          
                          self.pcmConverter?.convert(to: output, error: nil) { packets, status in
                              status.pointee = .haveData
                              return self.pcmInputBuffer    // I know this is wrong, but i'm not sure how to do it correctly
                          }
                          try audioEngine.start()
                      }
                  }
              }
          }
      }
  }

Answer 1

推测，不正确的答案

pcmConverter?.channelMap = [0, 0]怎么样？

实际答案

您不需要使用音频转换器通道映射，因为mono到立体声AVAudioConverters在默认情况下似乎复制了单声道。主要问题是outputFrameCapacity错了，在调用audioEngine.prepare()或启动引擎之前使用mainMixer的outputFormat。

假设sampleRate = 8000，修改后的解决方案如下所示：

private func loadAudioData(audioData: Data?) throws  {
    // Load audio data into player
    
    guard let audio = audioData else {return}
    do {
        audioEngine.attach(playerNode)
        audioEngine.connect(playerNode, to: audioEngine.mainMixerNode, format: nil)
        audioEngine.prepare() // https://stackoverflow.com/a/70392017/22147
        
        let outputAudioFormat = self.audioEngine.mainMixerNode.outputFormat(forBus: 0)
        guard let inputAudioFormat = AVAudioFormat(commonFormat: .pcmFormatInt16, sampleRate: Double(sampleRate), channels: 1, interleaved: false) else { return }
        
        let inputStreamDescription = inputAudioFormat.streamDescription.pointee
        let outputStreamDescription = outputAudioFormat.streamDescription.pointee
        let count = UInt32(audio.count)
        if count > 0 {
            let ibpf = inputStreamDescription.mBytesPerFrame
            let inputFrameCapacity = count / ibpf
            let outputFrameCapacity = Float64(inputFrameCapacity) * outputStreamDescription.mSampleRate / inputStreamDescription.mSampleRate
            self.pcmInputBuffer = AVAudioPCMBuffer(pcmFormat: inputAudioFormat, frameCapacity: inputFrameCapacity)
            self.pcmOutputBuffer = AVAudioPCMBuffer(pcmFormat: outputAudioFormat, frameCapacity: AVAudioFrameCount(outputFrameCapacity))
            
            if let input = self.pcmInputBuffer, let output = self.pcmOutputBuffer {
                self.pcmConverter = AVAudioConverter(from: inputAudioFormat, to: outputAudioFormat)
                input.frameLength = input.frameCapacity
                
                let b = UnsafeMutableBufferPointer(start: input.int16ChannelData?[0], count: input.stride * Int(inputFrameCapacity))
                let bytesCopied = audio.copyBytes(to: b)
                assert(bytesCopied == count)
                
                self.pcmConverter?.convert(to: output, error: nil) { packets, status in
                    status.pointee = .haveData
                    return self.pcmInputBuffer    // I know this is wrong, but i'm not sure how to do it correctly
                }
                try audioEngine.start()
                
                self.playerNode.scheduleBuffer(output, completionHandler: nil)
                self.playerNode.play()
            }
        }
    }
}