NAudio - wave split and combine wave in realtime

Question

I am working with a multi input soundcard and I want to achieve live mixing of multiple inputs. All the inputs are stereo, so I need to split them in first place, mix a selection of channel and provide them as mono stream.

The goal would be something like this mix Channel1[left] + Channel3[right] + Channel4[right] -> mono stream.

I have already implemented a process chain like this:

1) WaveIn -> create BufferedWaveProvider for each channel -> add Samples (just the ones for current channel) to each BufferedWaveProvider by using wavein.DataAvailable += { buffwavprovider[channel].AddSamples(...)... This gives me a nice list of multiple BufferdWaveProvider. The splitting audio part here is implemented correctly.

2) Select multiple BufferedWaveProviders and give them to MixingWaveProvider32. Then create a WaveStream (using WaveMixerStream32 and IWaveProvider).

3) A MultiChannelToMonoStream takes that WaveStream and generates a mixdown. This also works.

But result is, that audio is chopped. Like some trouble with the buffer....

Is this the correct way to handle this problem, or is there a way better solution around?

edit - code added:

public class AudioSplitter
   {
      public List<NamedBufferedWaveProvider> WaveProviders { private set; get; }
      public string Name { private set; get; }
      private WaveIn _wavIn;
      private int bytes_per_sample = 4;

      /// <summary>
      /// Splits up one WaveIn into one BufferedWaveProvider for each channel
      /// </summary>
      /// <param name="wavein"></param>
      /// <returns></returns>
      public AudioSplitter(WaveIn wavein, string name)
      {
         if (wavein.WaveFormat.Encoding != WaveFormatEncoding.IeeeFloat)
            throw new Exception("Format must be IEEE float");


         WaveProviders = new List<NamedBufferedWaveProvider>(wavein.WaveFormat.Channels);

         Name = name;
         _wavIn = wavein;
         _wavIn.StartRecording();
         var outFormat = NAudio.Wave.WaveFormat.CreateIeeeFloatWaveFormat(wavein.WaveFormat.SampleRate, 1);

         for (int i = 0; i < wavein.WaveFormat.Channels; i++)
         {
            WaveProviders.Add(new NamedBufferedWaveProvider(outFormat) { DiscardOnBufferOverflow = true, Name = Name + "_" + i });
         }

         bytes_per_sample = _wavIn.WaveFormat.BitsPerSample / 8;
         wavein.DataAvailable += Wavein_DataAvailable;
      }


      /// <summary>
      /// add samples for each channel to bufferedwaveprovider
      /// </summary>
      /// <param name="sender"></param>
      /// <param name="e"></param>
      private void Wavein_DataAvailable(object sender, WaveInEventArgs e)
      {
         int channel = 0;
         byte[] buffer = e.Buffer;
         for (int i = 0; i < e.BytesRecorded - bytes_per_sample; i = i + bytes_per_sample)
         {
            byte[] channel_buffer = new byte[bytes_per_sample];

            for (int j = 0; j < bytes_per_sample; j++)
            {
               channel_buffer[j] = buffer[i + j];
            }

            WaveProviders[channel].AddSamples(channel_buffer, 0, channel_buffer.Length);

            channel++;

            if (channel >= _wavIn.WaveFormat.Channels)
               channel = 0;

         }

      }
   }

Using the Audiosplitter for each channel gives a list of buffered wave provider (mono 32bit float).

 var mix = new MixingWaveProvider32(_waveProviders);
 var wps = new WaveProviderToWaveStream(mix);
 MultiChannelToMonoStream mms = new MultiChannelToMonoStream(wps);

 new Thread(() =>
  {
     byte[] buffer = new byte[4096];

     while (mms.Read(buffer, 0, buffer.Length) > 0 && isrunning)
     {


        using (FileStream fs = new FileStream("C:\\temp\\audio\\mono_32.wav", FileMode.Append, FileAccess.Write))
        {
           fs.Write(buffer, 0, buffer.Length);
        }

     }
  }).Start();

Answer 1

there is some space left for optimization, but basically this gets the job done:

  private void Wavein_DataAvailable(object sender, WaveInEventArgs e)
      {
         int channel = 0;
         byte[] buffer = e.Buffer;

         List<List<byte>> channelbuffers = new List<List<byte>>();
         for (int c = 0; c < _wavIn.WaveFormat.Channels; c++)
         {
            channelbuffers.Add(new List<byte>());
         }

         for (int i = 0; i < e.BytesRecorded; i++)
         {
            var byteList = channelbuffers[channel];

            byteList.Add(buffer[i]);

            if (i % bytes_per_sample == bytes_per_sample - 1)
               channel++;

            if (channel >= _wavIn.WaveFormat.Channels)
               channel = 0;
         }

         for (int j = 0; j < channelbuffers.Count; j++)
         {
            WaveProviders[j].AddSamples(channelbuffers[j].ToArray(), 0, channelbuffers[j].Count());
         }

      }

We need to provide a WaveProvider (WaveProviders[j]) for each channel.