public AudioContext GetAudioContext()
{
var resampler = new ResampleFilter(rawAudioFormat, transmittedAudioFormat);
var conferenceDtx = new DtxFilter(transmittedAudioFormat);
IAudioTwoWayFilter enhancer = null;
switch (enhancementStack)
{
case SpeechEnhancementStack.None:
enhancer = new NullEchoCancelFilter(mediaConfig.ExpectedAudioLatency, mediaConfig.FilterLength, transmittedAudioFormat, AudioFormat.Default);
break;
case SpeechEnhancementStack.Speex:
enhancer = new SpeexEchoCanceller2(mediaConfig, transmittedAudioFormat, AudioFormat.Default);
break;
case SpeechEnhancementStack.WebRtc:
enhancer = new WebRtcFilter(mediaConfig.ExpectedAudioLatency, mediaConfig.FilterLength, transmittedAudioFormat, AudioFormat.Default, mediaConfig.EnableAec, mediaConfig.EnableDenoise, mediaConfig.EnableAgc);
break;
}
IAudioEncoder encoder = null;
switch (codecType)
{
case AudioCodecType.G711M:
encoder = new G711MuLawEncoder(transmittedAudioFormat);
break;
case AudioCodecType.Speex:
encoder = new SpeexEncoder(transmittedAudioFormat);
break;
}
var ctx = new AudioContext(transmittedAudioFormat, resampler, conferenceDtx, enhancer, encoder);
return(ctx);
}
/// <summary>
/// Creates a new instance of the AudioContextFactory.
/// </summary>
/// <param name="rawAudioFormat">The format in which the audio coming directly from the microphone is recorded</param>
/// <param name="playedAudioFormat">The format in which the audio will be played back on the far end (typically 16Khz)</param>
/// <param name="config">The currently active MediaConfig instance</param>
/// <param name="mediaEnvironment">An IMediaEnvironment instance which can be used to make decisions about which context to return, for instance,
/// if the CPU is running too hot, or multiple people have joined the conference.</param>
public AudioContextFactory(AudioFormat rawAudioFormat, AudioFormat playedAudioFormat, MediaConfig config, IMediaEnvironment mediaEnvironment)
{
RawAudioFormat = rawAudioFormat;
PlayedAudioFormat = playedAudioFormat;
MediaConfig = config;
MediaEnvironment = mediaEnvironment;
// What we should use when there's only one other person, and CPU is OK:
// 16Khz, Speex, WebRtc at full strength
var directAudioFormat = new AudioFormat();
var directResampler = new ResampleFilter(rawAudioFormat, directAudioFormat);
directResampler.InstanceName = "High Quality Direct Resampler";
var directEnhancer = new WebRtcFilter(config.ExpectedAudioLatency, config.FilterLength, directAudioFormat, playedAudioFormat, config.EnableAec, config.EnableDenoise, config.EnableAgc);
directEnhancer.InstanceName = "High";
var directDtx = new DtxFilter(directAudioFormat);
var directEncoder = new SpeexEncoder(directAudioFormat);
HighQualityDirectCtx = new AudioContext(directAudioFormat, directResampler, directDtx, directEnhancer, directEncoder);
HighQualityDirectCtx.Description = "High Quality Direct";
// What we should use when there are multiple people (and hence the audio will need to be decoded and mixed), but CPU is OK:
// 8Khz, G711, WebRtc at full strength
var conferenceAudioFormat = new AudioFormat(AudioConstants.NarrowbandSamplesPerSecond);
var conferenceResampler = new ResampleFilter(rawAudioFormat, conferenceAudioFormat);
conferenceResampler.InstanceName = "High Quality Conference Resampler";
var conferenceEnhancer = new WebRtcFilter(config.ExpectedAudioLatency, config.FilterLength, conferenceAudioFormat, playedAudioFormat, config.EnableAec, config.EnableDenoise, config.EnableAgc);
conferenceEnhancer.InstanceName = "Medium";
var conferenceDtx = new DtxFilter(conferenceAudioFormat);
var conferenceEncoder = new G711MuLawEncoder(conferenceAudioFormat);
HighQualityConferenceCtx = new AudioContext(conferenceAudioFormat, conferenceResampler, conferenceDtx, conferenceEnhancer, conferenceEncoder);
HighQualityConferenceCtx.Description = "High Quality Conference";
// What we should use when one or more remote CPU's isn't keeping up (regardless of how many people are in the room):
// 8Khz, G711, WebRtc at full-strength
var remoteFallbackAudioFormat = new AudioFormat(AudioConstants.NarrowbandSamplesPerSecond);
var remoteFallbackResampler = new ResampleFilter(rawAudioFormat, remoteFallbackAudioFormat);
remoteFallbackResampler.InstanceName = "Low Quality Remote CPU Resampler";
var remoteFallbackEnhancer = new WebRtcFilter(config.ExpectedAudioLatency, config.FilterLength, remoteFallbackAudioFormat, playedAudioFormat, config.EnableAec, config.EnableDenoise, config.EnableAgc);
remoteFallbackEnhancer.InstanceName = "Medium";
var remoteFallbackDtx = new DtxFilter(remoteFallbackAudioFormat);
var remoteFallbackEncoder = new G711MuLawEncoder(remoteFallbackAudioFormat);
LowQualityForRemoteCpuCtx = new AudioContext(remoteFallbackAudioFormat, remoteFallbackResampler, remoteFallbackDtx, remoteFallbackEnhancer, remoteFallbackEncoder);
LowQualityForRemoteCpuCtx.Description = "Fallback for remote high CPU";
// What we should use when the local CPU isn't keeping up (regardless of how many people are in the room):
// 8Khz, G711, WebRtc at half-strength
var fallbackAudioFormat = new AudioFormat(AudioConstants.NarrowbandSamplesPerSecond);
var fallbackResampler = new ResampleFilter(rawAudioFormat, fallbackAudioFormat);
fallbackResampler.InstanceName = "Low Quality Local CPU Resampler";
var fallbackEnhancer = new WebRtcFilter(config.ExpectedAudioLatencyFallback, config.FilterLengthFallback, fallbackAudioFormat, playedAudioFormat, config.EnableAec, false, false);
fallbackEnhancer.InstanceName = "Low";
var fallbackDtx = new DtxFilter(fallbackAudioFormat);
var fallbackEncoder = new G711MuLawEncoder(fallbackAudioFormat);
LowQualityForLocalCpuCtx = new AudioContext(fallbackAudioFormat, fallbackResampler, fallbackDtx, fallbackEnhancer, fallbackEncoder);
LowQualityForLocalCpuCtx.Description = "Fallback for local high CPU";
_audioContextAdapter = new EnvironmentAdapter <AudioContext>(mediaEnvironment,
HighQualityDirectCtx,
HighQualityConferenceCtx,
LowQualityForRemoteCpuCtx,
LowQualityForLocalCpuCtx);
}
public void CreateAudioContexts()
{
_captureSource.VideoCaptureDevice = null;
if (_captureSource.AudioCaptureDevice == null)
{
_captureSource.AudioCaptureDevice = CaptureDeviceConfiguration.GetDefaultAudioCaptureDevice();
if (_captureSource.AudioCaptureDevice == null)
{
throw new InvalidOperationException("No suitable audio capture device was found");
}
}
MediaDeviceConfig.SelectBestAudioFormat(_captureSource.AudioCaptureDevice);
_captureSource.AudioCaptureDevice.AudioFrameSize = AudioFormat.Default.MillisecondsPerFrame; // 20 milliseconds
var desiredFormat = _captureSource.AudioCaptureDevice.DesiredFormat;
var rawAudioFormat = new AudioFormat(desiredFormat.SamplesPerSecond, AudioFormat.Default.MillisecondsPerFrame, desiredFormat.Channels, desiredFormat.BitsPerSample);
var playedAudioFormat = new AudioFormat();
var config = MediaConfig.Default;
// Absolutely bare minimum processing - doesn't process sound at all.
var nullAudioFormat = new AudioFormat();
var nullResampler = new ResampleFilter(rawAudioFormat, nullAudioFormat);
nullResampler.InstanceName = "Null resample filter";
var nullEnhancer = new NullEchoCancelFilter(config.ExpectedAudioLatency, config.FilterLength, nullAudioFormat, playedAudioFormat);
nullEnhancer.InstanceName = "Null";
var nullDtx = new NullAudioInplaceFilter();
var nullEncoder = new NullAudioEncoder();
var nullAudioContext = new AudioContext(nullAudioFormat, nullResampler, nullDtx, nullEnhancer, nullEncoder);
nullAudioContext.Description = "Null";
// What we should use when there's only one other person, and CPU is OK:
// 16Khz, Speex, WebRtc at full strength
var directAudioFormat = new AudioFormat();
var directResampler = new ResampleFilter(rawAudioFormat, directAudioFormat);
directResampler.InstanceName = "Direct high quality resample filter";
var directEnhancer = new WebRtcFilter(config.ExpectedAudioLatency, config.FilterLength, directAudioFormat, playedAudioFormat, config.EnableAec, config.EnableDenoise, config.EnableAgc);
directEnhancer.InstanceName = "High";
var directDtx = new DtxFilter(directAudioFormat);
var directEncoder = new SpeexEncoder(directAudioFormat);
var directAudioContext = new AudioContext(directAudioFormat, directResampler, directDtx, directEnhancer, directEncoder);
directAudioContext.Description = "High Quality Direct";
// What we should use when there are multiple people (and hence the audio will need to be decoded and mixed), but CPU is OK:
// 8Khz, G711, WebRtc at full strength
var conferenceAudioFormat = new AudioFormat(AudioConstants.NarrowbandSamplesPerSecond);
var conferenceResampler = new ResampleFilter(rawAudioFormat, conferenceAudioFormat);
conferenceResampler.InstanceName = "Conference high quality resample filter";
var conferenceEnhancer = new WebRtcFilter(config.ExpectedAudioLatency, config.FilterLength, conferenceAudioFormat, playedAudioFormat, config.EnableAec, config.EnableDenoise, config.EnableAgc);
conferenceEnhancer.InstanceName = "Medium";
var conferenceDtx = new DtxFilter(conferenceAudioFormat);
var conferenceEncoder = new G711MuLawEncoder(conferenceAudioFormat);
var conferenceAudioContext = new AudioContext(conferenceAudioFormat, conferenceResampler, conferenceDtx, conferenceEnhancer, conferenceEncoder);
conferenceAudioContext.Description = "High Quality Conference";
// What we should use when one or more remote CPU's isn't keeping up (regardless of how many people are in the room):
// 8Khz, G711, WebRtc at full-strength
var remoteFallbackAudioFormat = new AudioFormat(AudioConstants.NarrowbandSamplesPerSecond);
var remoteFallbackResampler = new ResampleFilter(rawAudioFormat, remoteFallbackAudioFormat);
remoteFallbackResampler.InstanceName = "Fallback remote high cpu resample filter";
var remoteFallbackEnhancer = new WebRtcFilter(config.ExpectedAudioLatency, config.FilterLength, remoteFallbackAudioFormat, playedAudioFormat, config.EnableAec, config.EnableDenoise, config.EnableAgc);
remoteFallbackEnhancer.InstanceName = "Medium";
var remoteFallbackDtx = new DtxFilter(remoteFallbackAudioFormat);
var remoteFallbackEncoder = new G711MuLawEncoder(remoteFallbackAudioFormat);
var remoteFallbackAudioContext = new AudioContext(remoteFallbackAudioFormat, remoteFallbackResampler, remoteFallbackDtx, remoteFallbackEnhancer, remoteFallbackEncoder);
remoteFallbackAudioContext.Description = "Fallback for remote high CPU";
// What we should use when the local CPU isn't keeping up (regardless of how many people are in the room):
// 8Khz, G711, WebRtc at half-strength
var fallbackAudioFormat = new AudioFormat(AudioConstants.NarrowbandSamplesPerSecond);
var fallbackResampler = new ResampleFilter(rawAudioFormat, fallbackAudioFormat);
fallbackResampler.InstanceName = "Fallback resample filter";
var fallbackEnhancer = new WebRtcFilter(config.ExpectedAudioLatencyFallback, config.FilterLengthFallback, fallbackAudioFormat, playedAudioFormat, config.EnableAec, false, false);
fallbackEnhancer.InstanceName = "Low";
var fallbackDtx = new DtxFilter(fallbackAudioFormat);
var fallbackEncoder = new G711MuLawEncoder(fallbackAudioFormat);
var fallbackAudioContext = new AudioContext(fallbackAudioFormat, fallbackResampler, fallbackDtx, fallbackEnhancer, fallbackEncoder);
fallbackAudioContext.Description = "Fallback for local high CPU";
AudioContextCollection.Clear();
AudioContextCollection.Add(nullAudioContext);
AudioContextCollection.Add(directAudioContext);
AudioContextCollection.Add(conferenceAudioContext);
AudioContextCollection.Add(remoteFallbackAudioContext);
AudioContextCollection.Add(fallbackAudioContext);
CurrentAudioContext = nullAudioContext;
}
