/// <summary>
///
/// </summary>
/// <param name="inputFormat">format of audio supplied to `Send`</param>
/// <param name="inputFrameSize">Size of frames which will should be provided from `GetFrameBuffer`</param>
/// <param name="intermediateFrameSize">Size of frames which should be passed into `PreprocessAudioFrame`</param>
/// <param name="intermediateSampleRate">Sample rate which should be passed into `PreprocessAudioFrame`</param>
/// <param name="outputFrameSize">Size of frames which will be provided sent to `SendSamplesToSubscribers`</param>
/// <param name="outputSampleRate"></param>
protected BasePreprocessingPipeline(WaveFormat inputFormat, int inputFrameSize, int intermediateFrameSize, int intermediateSampleRate, int outputFrameSize, int outputSampleRate)
{
if (inputFrameSize < 0)
{
throw new ArgumentOutOfRangeException("inputFrameSize", "Input frame size cannot be less than zero");
}
_inputFrameSize = inputFrameSize;
_outputFrameSize = outputFrameSize;
_outputFormat = new WaveFormat(1, outputSampleRate);
//Create input system (source of empty buffers, queue of waiting input data)
_inputBufferSource = new ConcurrentPool <float[]>(24, () => new float[inputFrameSize]);
_inputQueue = new TransferBuffer <float[]>(12);
_emptyInputFrame = new float[inputFrameSize];
//Create resampler to resample input to intermediate rate
_resamplerInput = new BufferedSampleProvider(inputFormat, InputFrameSize * 4);
_resampler = new Resampler(_resamplerInput, 48000);
_resampledOutput = new SampleToFrameProvider(_resampler, (uint)OutputFrameSize);
_intermediateFrame = new float[intermediateFrameSize];
_threadEvent = new AutoResetEvent(false);
_thread = new DThread(ThreadEntry);
}
public EncoderPipeline([NotNull] WaveFormat inputFormat, [NotNull] IVoiceEncoder encoder, [NotNull] ICommsNetwork net)
{
if (inputFormat == null)
{
throw new ArgumentNullException("inputFormat");
}
if (encoder == null)
{
throw new ArgumentNullException("encoder");
}
if (net == null)
{
throw new ArgumentNullException("net");
}
_net = net;
_inputFormat = inputFormat;
_encoder = new ReadonlyLockedValue <IVoiceEncoder>(encoder);
//Create buffers to store the encoder input (1 frame of floats) and output (twice equivalent amount of bytes)
_plainSamples = new float[encoder.FrameSize];
_encodedBytes = new byte[encoder.FrameSize * sizeof(float) * 2];
//Input buffer to store raw data from microphone
_input = new BufferedSampleProvider(_inputFormat, encoder.FrameSize * 2);
//Resample data from microphone rate -> encoder rate
_resampler = new Resampler(_input, encoder.SampleRate);
//Provides encoder sized and encoder rate frames of data
_output = new SampleToFrameProvider(_resampler, (uint)encoder.FrameSize);
}
public EncoderPipeline(IMicrophoneCapture mic, IVoiceEncoder encoder, ICommsNetwork net, Func <int> channelCount)
{
_mic = mic;
_encoder = encoder;
_net = net;
_channelCount = channelCount;
_encodedBytes = new byte[encoder.FrameSize * sizeof(float)];
_plainSamples = new float[encoder.FrameSize];
_inputFormat = mic.Format;
//Create an input buffer with plenty of spare space
_input = new BufferedSampleProvider(_inputFormat, Math.Max(_encoder.FrameSize * 2, mic.FrameSize * 2));
_resampler = new Resampler(_input, _encoder.SampleRate);
//Whatever we did above, we need to read in frame size chunks
_output = new SampleToFrameProvider(_resampler, (uint)encoder.FrameSize);
}
/// <summary>
///
/// </summary>
/// <param name="inputFormat">format of audio supplied to `Send`</param>
/// <param name="intermediateFrameSize">Size of frames which should be passed into `PreprocessAudioFrame`</param>
/// <param name="intermediateSampleRate">Sample rate which should be passed into `PreprocessAudioFrame`</param>
/// <param name="outputFrameSize">Size of frames which will be provided sent to `SendSamplesToSubscribers`</param>
/// <param name="outputSampleRate"></param>
protected BasePreprocessingPipeline([NotNull] WaveFormat inputFormat, int intermediateFrameSize, int intermediateSampleRate, int outputFrameSize, int outputSampleRate)
{
if (inputFormat == null)
{
throw new ArgumentNullException("inputFormat");
}
if (intermediateFrameSize < 0)
{
throw new ArgumentOutOfRangeException("intermediateFrameSize", "Intermediate frame size cannot be less than zero");
}
if (intermediateSampleRate < 0)
{
throw new ArgumentOutOfRangeException("intermediateSampleRate", "Intermediate sample rate cannot be less than zero");
}
if (outputFrameSize < 0)
{
throw new ArgumentOutOfRangeException("outputFrameSize", "Output frame size cannot be less than zero");
}
if (outputSampleRate < 0)
{
throw new ArgumentOutOfRangeException("outputSampleRate", "Output sample rate cannot be less than zero");
}
_outputFrameSize = outputFrameSize;
_outputFormat = new WaveFormat(outputSampleRate, 1);
//Create resampler to resample input to intermediate rate
_resamplerInput = new BufferedSampleProvider(inputFormat, intermediateFrameSize * 16);
_resampler = new Resampler(_resamplerInput, 48000);
_resampledOutput = new SampleToFrameProvider(_resampler, (uint)OutputFrameSize);
_intermediateFrame = new float[intermediateFrameSize];
//Create a thread to drive the audio processing
_threadEvent = new AutoResetEvent(false);
_thread = new DThread(ThreadEntry);
// We don't want to overestimate the latency. It's hard to come up with a reasonable number for this because if the input frames are >= the...
// ...intermediate frames there will actually be no buffering delay in the preprocessor (it'll pick up a complete frame and process it right away).
_estimatedPreprocessorLatencyMs = 0;
}
/// <summary>
///
/// </summary>
/// <param name="inputFormat">format of audio supplied to `Send`</param>
/// <param name="inputFrameSize">Size of frames which will should be provided from `GetFrameBuffer`</param>
/// <param name="intermediateFrameSize">Size of frames which should be passed into `PreprocessAudioFrame`</param>
/// <param name="intermediateSampleRate">Sample rate which should be passed into `PreprocessAudioFrame`</param>
/// <param name="outputFrameSize">Size of frames which will be provided sent to `SendSamplesToSubscribers`</param>
/// <param name="outputSampleRate"></param>
protected BasePreprocessingPipeline([NotNull] WaveFormat inputFormat, int inputFrameSize, int intermediateFrameSize, int intermediateSampleRate, int outputFrameSize, int outputSampleRate)
{
if (inputFormat == null)
{
throw new ArgumentNullException("inputFormat");
}
if (inputFrameSize < 0)
{
throw new ArgumentOutOfRangeException("inputFrameSize", "Input frame size cannot be less than zero");
}
if (intermediateFrameSize < 0)
{
throw new ArgumentOutOfRangeException("intermediateFrameSize", "Intermediate frame size cannot be less than zero");
}
if (intermediateSampleRate < 0)
{
throw new ArgumentOutOfRangeException("intermediateSampleRate", "Intermediate sample rate cannot be less than zero");
}
if (outputFrameSize < 0)
{
throw new ArgumentOutOfRangeException("outputFrameSize", "Output frame size cannot be less than zero");
}
if (outputSampleRate < 0)
{
throw new ArgumentOutOfRangeException("outputSampleRate", "Output sample rate cannot be less than zero");
}
_inputFrameSize = inputFrameSize;
_outputFrameSize = outputFrameSize;
_outputFormat = new WaveFormat(1, outputSampleRate);
//Create resampler to resample input to intermediate rate
_resamplerInput = new BufferedSampleProvider(inputFormat, InputFrameSize * 16);
_resampler = new Resampler(_resamplerInput, 48000);
_resampledOutput = new SampleToFrameProvider(_resampler, (uint)OutputFrameSize);
_intermediateFrame = new float[intermediateFrameSize];
_threadEvent = new AutoResetEvent(false);
_thread = new DThread(ThreadEntry);
}
/// <param name="micName"></param>
/// <param name="source">Source to read frames from</param>
private MicrophoneCapture([CanBeNull] string micName, AudioClip source)
{
if (source == null)
{
throw new ArgumentNullException("source", Log.PossibleBugMessage("capture source clip is null", "333E11A6-8026-41EB-9B34-EF9ADC54B651"));
}
_micName = micName;
_clip = source;
var captureFormat = new WaveFormat(1, source.frequency);
_maxReadBufferPower = (byte)Math.Ceiling(Math.Log(0.1f * source.frequency, 2));
_preprocessing = new WebRtcPreprocessingPipeline(captureFormat);
_preprocessing.Start();
//Ensure we have enough buffer size to contain several input frames to the preprocessor
_rawMicSamples = new BufferedSampleProvider(captureFormat, _preprocessing.InputFrameSize * 4);
_rawMicFrames = new SampleToFrameProvider(_rawMicSamples, (uint)_preprocessing.InputFrameSize);
Log.Info("Began mic capture (SampleRate:{0} FrameSize:{1}, Buffer Limit:2^{2})", captureFormat.SampleRate, _preprocessing.InputFrameSize, _maxReadBufferPower);
}
public virtual WaveFormat StartCapture(string inputMicName)
{
//Sanity checks
Log.AssertAndThrowPossibleBug(_clip == null, "1BAD3E74-B451-4B7D-A9B9-35225BE55364", "Attempted to Start microphone capture, but capture is already running");
//Early exit if there are no microphones connected
if (Log.AssertAndLogWarn(Microphone.devices.Length > 0, "No microphone detected; disabling voice capture"))
{
return(null);
}
//Check the micName and default to null if it's invalid (all whitespace or not a known device)
_micName = ChooseMicName(inputMicName);
//Get device capabilities and choose a sample rate as close to 48000 as possible.
//If min and max are both zero that indicates we can use any sample rate
int minFreq;
int maxFreq;
Microphone.GetDeviceCaps(_micName, out minFreq, out maxFreq);
var sampleRate = minFreq == 0 && maxFreq == 0 ? 48000 : Mathf.Clamp(48000, minFreq, maxFreq);
Log.Debug("GetDeviceCaps name=`{0}` min=`{1}` max=`{2}`", _micName, minFreq, maxFreq);
//Get the audioclip from Unity for this microphone (with a fairly large internal buffer)
_clip = Microphone.Start(_micName, true, 10, sampleRate);
if (_clip == null)
{
Log.Error("Failed to start microphone capture");
return(null);
}
//Setup buffers for capture
_format = new WaveFormat(_clip.frequency, 1);
_maxReadBufferPower = (byte)Math.Ceiling(Math.Log(0.1f * _clip.frequency, 2));
// Create/resize the audio buffers to contain 20ms frames of data. Any frame size will work (the pipeline will buffer/split them as necessary) but 20ms is
// optimal because that's native frame size the preprocessor works at so it has to do no extra work to assemble the frames at it's desired size.
var frameSize = (int)(0.02 * _clip.frequency);
if (_rawMicSamples == null || _rawMicSamples.WaveFormat != _format || _rawMicSamples.Capacity != frameSize || _rawMicFrames.FrameSize != frameSize)
{
_rawMicSamples = new BufferedSampleProvider(_format, frameSize * 4);
_rawMicFrames = new SampleToFrameProvider(_rawMicSamples, (uint)frameSize);
}
if (_frame == null || _frame.Length != frameSize)
{
_frame = new float[frameSize];
}
//watch for device changes - we need to reset if the audio device changes
AudioSettings.OnAudioConfigurationChanged += OnAudioDeviceChanged;
_audioDeviceChanged = false;
//Reset subscribers to prepare them for another stream of data
for (var i = 0; i < _subscribers.Count; i++)
{
_subscribers[i].Reset();
}
Latency = TimeSpan.FromSeconds(frameSize / (float)_format.SampleRate);
Log.Info("Began mic capture (SampleRate:{0}Hz, FrameSize:{1}, Buffer Limit:2^{2}, Latency:{3}ms, Device:'{4}')", _clip.frequency, frameSize, _maxReadBufferPower, Latency.TotalMilliseconds, _micName);
return(_format);
}