/// <summary>
/// Initialize the capturer.
/// </summary>
/// <param name="engineLatency">
/// Number of milliseconds of acceptable lag between live sound being produced and recording operation.
/// </param>
/// <param name="gain">
/// The gain to be applied to the audio after capture.
/// </param>
/// <param name="outFormat">
/// The format of the audio to be captured. If this is NULL, the default audio format of the
/// capture device will be used.
/// </param>
/// <param name="callback">
/// Callback function delegate which will handle the captured data.
/// </param>
/// <param name="speech">
/// If true, sets the audio category to speech to optimize audio pipeline for speech recognition.
/// </param>
public void Initialize(int engineLatency, float gain, WaveFormat outFormat, AudioDataAvailableCallback callback, bool speech)
{
// Create our shutdown event - we want a manual reset event that starts in the not-signaled state.
this.shutdownEvent = new ManualResetEvent(false);
// Now activate an IAudioClient object on our preferred endpoint and retrieve the mix format for that endpoint.
object obj = this.endpoint.Activate(ref audioClientIID, ClsCtx.INPROC_SERVER, IntPtr.Zero);
this.audioClient = (IAudioClient)obj;
// The following block enables advanced mic array APO pipeline on Windows 10 RS2 builds >= 15004.
// This must be called before the call to GetMixFormat() in LoadFormat().
if (speech)
{
IAudioClient2 audioClient2 = (IAudioClient2)this.audioClient;
if (audioClient2 != null)
{
AudioClientProperties properties = new AudioClientProperties
{
Size = Marshal.SizeOf <AudioClientProperties>(),
Category = AudioStreamCategory.Speech
};
int hr = audioClient2.SetClientProperties(ref properties);
if (hr != 0)
{
Console.WriteLine("Failed to set audio stream category to AudioCategory_Speech: {0}", hr);
}
}
else
{
Console.WriteLine("Unable to get IAudioClient2 interface");
}
}
// Load the MixFormat. This may differ depending on the shared mode used.
this.LoadFormat();
// Remember our configured latency
this.engineLatencyInMs = engineLatency;
// Set the gain
this.gain = gain;
// Determine whether or not we need a resampler
this.resampler = null;
if (outFormat != null)
{
// Check if the desired format is supported
IntPtr closestMatchPtr;
IntPtr outFormatPtr = WaveFormat.MarshalToPtr(outFormat);
int hr = this.audioClient.IsFormatSupported(AudioClientShareMode.Shared, outFormatPtr, out closestMatchPtr);
// Free outFormatPtr to prevent leaking memory
Marshal.FreeHGlobal(outFormatPtr);
if (hr == 0)
{
// Replace _MixFormat with outFormat. Since it is supported, we will initialize
// the audio capture client with that format and capture without resampling.
this.mixFormat = outFormat;
this.mixFrameSize = (this.mixFormat.BitsPerSample / 8) * this.mixFormat.Channels;
}
else
{
// In all other cases, we need to resample to OutFormat
if ((hr == 1) && (closestMatchPtr != IntPtr.Zero))
{
// Use closest match suggested by IsFormatSupported() and resample
this.mixFormat = WaveFormat.MarshalFromPtr(closestMatchPtr);
this.mixFrameSize = (this.mixFormat.BitsPerSample / 8) * this.mixFormat.Channels;
// Free closestMatchPtr to prevent leaking memory
Marshal.FreeCoTaskMem(closestMatchPtr);
}
this.inputBufferSize = (int)(this.engineLatencyInMs * this.mixFormat.AvgBytesPerSec / 1000);
this.outputBufferSize = (int)(this.engineLatencyInMs * outFormat.AvgBytesPerSec / 1000);
DeviceUtil.CreateResamplerBuffer(this.inputBufferSize, out this.inputSample, out this.inputBuffer);
DeviceUtil.CreateResamplerBuffer(this.outputBufferSize, out this.outputSample, out this.outputBuffer);
// Create resampler object
this.resampler = DeviceUtil.CreateResampler(this.mixFormat, outFormat);
}
}
this.InitializeAudioEngine();
// Set the callback function
this.dataAvailableCallback = callback;
}
private void InitializeCaptureDevice(IAudioClient2 audioClientInterface)
{
this.audioClient = new AudioClient2(audioClientInterface);
// Activation complete. Set the client properties
AudioClientProperties props = new AudioClientProperties();
props.cbSize = 16; // (uint)System.Runtime.InteropServices.Marshal.SizeOf<AudioClientProperties>();
props.bIsOffload = false; // FALSE
props.eCategory = AudioStreamCategory.Communications; // AUDIO_STREAM_CATEGORY::AudioCategory_Communications in C++
props.Options = AudioClientStreamOptions.None; // AUDCLNT_STREAMOPTIONS_NONE in C++
int hresult = audioClientInterface.SetClientProperties(props);
if (hresult != 0)
{
Marshal.ThrowExceptionForHR(hresult);
}
this.waveFormat = this.audioClient.MixFormat;
//if (this.isXMOS)
//{
// // the mix format for the XMOS is likely a 6 channel interleaved audio stream that we don't need.
// // in theory, we should be able to just request a single channel 48K stream and it will just work
// // and it will be good!!! Of course, this doesn't actually work... Set the veil audio mic to be 1 channel
// // before doing anything else...
// this.waveFormat = WaveFormatExtensible.CreateIeeeFloatWaveFormat(48000, 1);
//}
long requestedDuration = REFTIMES_PER_MILLISEC * 100;
this.frequency = 8 * waveFormat.AverageBytesPerSecond / (waveFormat.Channels * waveFormat.BitsPerSample);
this.EventWriterDLL.BuildLine(
"+2 start => WasapiCapture::InitializeCaptureDevice => Wave Format =\n" +
" => average bytes per second = " + waveFormat.AverageBytesPerSecond + "\n" +
" => bits per sample = " + waveFormat.BitsPerSample + "\n" +
" => channels = " + waveFormat.Channels + "\n" +
" => encoding = " + waveFormat.WaveFormatTag + "\n" +
" => extra size = " + waveFormat.ExtraSize + "\n" +
" => frequency = " + frequency);
hresult = this.audioClient.Initialize(AudioClientShareMode.Shared,
AudioClientStreamFlags.EventCallback,
requestedDuration,
0,
ref this.waveFormat,
Guid.Empty);
if (hresult == 0)
{
int bufferFrameCount = this.audioClient.BufferSize;
this.bytesPerFrame = this.waveFormat.Channels * this.waveFormat.BitsPerSample / 8;
this.recordBuffer = new byte[bufferFrameCount * bytesPerFrame];
this.EventWriterDLL.BuildLine(
"+3 => WasapiCapture::InitializeCaptureDevice => " + string.Format("record buffer size = {0}", this.recordBuffer.Length));
// Get back the effective latency from AudioClient
this.latencyMilliseconds = (int)(this.audioClient.StreamLatency / 10000);
this.initialized = true;
}
else
{
this.EventWriterDLL.BuildLine("-3 => WasapiCapture::InitializeCaptureDevice => Error:" + string.Format("{0:X}", hresult));
}
}
