The Microsoft audio recording example shows how to record audio, but it just gets a PCM encoded stream of samples – if you want to send this audio stream off somewhere else, to be played later, it needs a proper header.

Below I show you how to use a couple of methods I’ve written (WriteWavHeader and UpdateWavHeader), and I include their implementation.

Assuming you start with the Microsoft sample, when you initiate the audio recording you initialize a memory stream into which the samples will be written:

private void StartRecording()
{
    _microphone.BufferDuration = TimeSpan.FromMilliseconds(500);

    // Allocate memory to hold the audio data
    _recordingBuffer = new byte[_microphone.GetSampleSizeInBytes(_microphone.BufferDuration)];

    _recordingStream.SetLength(0);

    WriteWavHeader(_recordingStream, _microphone.SampleRate);

    _microphone.Start();
}

Note the call to WriteWavHeader to write the wav header to the start of the audio stream.

When the recording finishes, you update the header (because we need to fill in fields based on the data length):

private void StopRecording()
{
    _microphone.Stop();
    UpdateWavHeader(_recordingStream);
}

Here are the implementations of the WriteWavHeader and UpdateWavHeader methods.  The comments come from this web page describing the wav header file format.

public void WriteWavHeader(Stream stream, int sampleRate)
{
    const int bitsPerSample = 16;
    const int bytesPerSample = bitsPerSample / 8;
    var encoding = System.Text.Encoding.UTF8;

    // ChunkID Contains the letters "RIFF" in ASCII form (0x52494646 big-endian form).
    stream.Write(encoding.GetBytes("RIFF"), 0, 4);

    // NOTE this will be filled in later
    stream.Write(BitConverter.GetBytes(0), 0, 4);

    // Format Contains the letters "WAVE"(0x57415645 big-endian form).
    stream.Write(encoding.GetBytes("WAVE"), 0, 4);

    // Subchunk1ID Contains the letters "fmt " (0x666d7420 big-endian form).
    stream.Write(encoding.GetBytes("fmt "), 0, 4);

    // Subchunk1Size 16 for PCM.  This is the size of therest of the Subchunk which follows this number.
    stream.Write(BitConverter.GetBytes(16), 0, 4);

    // AudioFormat PCM = 1 (i.e. Linear quantization) Values other than 1 indicate some form of compression.
    stream.Write(BitConverter.GetBytes((short)1), 0, 2);

    // NumChannels Mono = 1, Stereo = 2, etc.
    stream.Write(BitConverter.GetBytes((short)1), 0, 2);

    // SampleRate 8000, 44100, etc.
    stream.Write(BitConverter.GetBytes(sampleRate), 0, 4);

    // ByteRate =  SampleRate * NumChannels * BitsPerSample/8
    stream.Write(BitConverter.GetBytes(sampleRate * bytesPerSample), 0, 4);

    // BlockAlign NumChannels * BitsPerSample/8 The number of bytes for one sample including all channels.
    stream.Write(BitConverter.GetBytes((short)(bytesPerSample)), 0, 2);

    // BitsPerSample    8 bits = 8, 16 bits = 16, etc.
    stream.Write(BitConverter.GetBytes((short)(bitsPerSample)), 0, 2);

    // Subchunk2ID Contains the letters "data" (0x64617461 big-endian form).
    stream.Write(encoding.GetBytes("data"), 0, 4);
    
    // NOTE to be filled in later
    stream.Write(BitConverter.GetBytes(0), 0, 4);
}

public void UpdateWavHeader(Stream stream)
{
    if (!stream.CanSeek) throw new Exception("Can&'t seek stream to update wav header");

    var oldPos = stream.Position;

    // ChunkSize  36 + SubChunk2Size
    stream.Seek(4, SeekOrigin.Begin);
    stream.Write(BitConverter.GetBytes((int)stream.Length - 8), 0, 4);

    // Subchunk2Size == NumSamples * NumChannels * BitsPerSample/8 This is the number of bytes in the data.
    stream.Seek(40, SeekOrigin.Begin);
    stream.Write(BitConverter.GetBytes((int)stream.Length - 44), 0, 4);

    stream.Seek(oldPos, SeekOrigin.Begin);
}