REFAC(client): Use AudioPreprocessor

src/mumble/AudioConfigDialog.cpp

@@ -586,7 +586,7 @@ void AudioInputDialog::showSpeexNoiseSuppressionSlider(bool show) {
 void AudioInputDialog::on_Tick_timeout() {
 	AudioInputPtr ai = Global::get().ai;
 
-	if (!ai.get() || !ai->sppPreprocess)
+	if (!ai.get() || !ai->m_preprocessor)
 		return;
 
 	abSpeech->iBelow = qsTransmitMin->value();

src/mumble/AudioInput.cpp

@@ -255,8 +255,7 @@ AudioInput::AudioInput()
 
 	bEchoMulti = false;
 
-	sppPreprocess = nullptr;
-	sesEcho       = nullptr;
+	sesEcho = nullptr;
 	srsMic = srsEcho = nullptr;
 
 	iEchoChannels = iMicChannels = 0;
@@ -298,8 +297,6 @@ AudioInput::~AudioInput() {
 	}
 #endif
 
-	if (sppPreprocess)
-		speex_preprocess_state_destroy(sppPreprocess);
 	if (sesEcho)
 		speex_echo_state_destroy(sesEcho);
 
@@ -740,44 +737,34 @@ void AudioInput::resetAudioProcessor() {
 	if (!bResetProcessor)
 		return;
 
-	int iArg;
-
-	if (sppPreprocess)
-		speex_preprocess_state_destroy(sppPreprocess);
 	if (sesEcho)
 		speex_echo_state_destroy(sesEcho);
 
-	sppPreprocess = speex_preprocess_state_init(iFrameSize, iSampleRate);
+	m_preprocessor.init(iSampleRate, iFrameSize);
 	resync.reset();
 	selectNoiseCancel();
 
-	iArg = 1;
-	speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_VAD, &iArg);
-	speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC, &iArg);
-	speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_DEREVERB, &iArg);
-
-	iArg = 30000;
-	speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC_TARGET, &iArg);
+	m_preprocessor.toggleVAD(true);
+	m_preprocessor.toggleAGC(true);
+	m_preprocessor.toggleDereverb(true);
 
-	float v = 30000.0f / static_cast< float >(Global::get().s.iMinLoudness);
-	iArg    = static_cast< int >(floorf(20.0f * log10f(v)));
-	speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC_MAX_GAIN, &iArg);
+	m_preprocessor.setAGCTarget(30000);
 
-	iArg = -60;
-	speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC_DECREMENT, &iArg);
+	const float v = 30000.0f / static_cast< float >(Global::get().s.iMinLoudness);
+	m_preprocessor.setAGCMaxGain(static_cast< std::int32_t >(floorf(20.0f * log10f(v))));
+	m_preprocessor.setAGCDecrement(-60);
 
 	if (noiseCancel == Settings::NoiseCancelSpeex) {
-		iArg = Global::get().s.iSpeexNoiseCancelStrength;
-		speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_NOISE_SUPPRESS, &iArg);
+		m_preprocessor.setNoiseSuppress(Global::get().s.iSpeexNoiseCancelStrength);
 	}
 
 	if (iEchoChannels > 0) {
 		int filterSize = iFrameSize * (10 + resync.getNominalLag());
 		sesEcho =
 			speex_echo_state_init_mc(iFrameSize, filterSize, 1, bEchoMulti ? static_cast< int >(iEchoChannels) : 1);
-		iArg = iSampleRate;
+		int iArg = iSampleRate;
 		speex_echo_ctl(sesEcho, SPEEX_ECHO_SET_SAMPLING_RATE, &iArg);
-		speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_ECHO_STATE, sesEcho);
+		m_preprocessor.setEchoState(sesEcho);
 
 		qWarning("AudioInput: ECHO CANCELLER ACTIVE");
 	} else {
@@ -821,24 +808,24 @@ void AudioInput::selectNoiseCancel() {
 #endif
 	}
 
-	int iArg = 0;
+	bool preprocessorDenoise = false;
 	switch (noiseCancel) {
 		case Settings::NoiseCancelOff:
 			qWarning("AudioInput: Noise canceller disabled");
 			break;
 		case Settings::NoiseCancelSpeex:
 			qWarning("AudioInput: Using Speex as noise canceller");
-			iArg = 1;
+			preprocessorDenoise = true;
 			break;
 		case Settings::NoiseCancelRNN:
 			qWarning("AudioInput: Using ReNameNoise as noise canceller");
 			break;
 		case Settings::NoiseCancelBoth:
-			iArg = 1;
+			preprocessorDenoise = true;
 			qWarning("AudioInput: Using ReNameNoise and Speex as noise canceller");
 			break;
 	}
-	speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_DENOISE, &iArg);
+	m_preprocessor.toggleDenoise(preprocessorDenoise);
 }
 
 int AudioInput::encodeOpusFrame(short *source, int size, EncodingOutputBuffer &buffer) {
@@ -857,7 +844,6 @@ int AudioInput::encodeOpusFrame(short *source, int size, EncodingOutputBuffer &b
 }
 
 void AudioInput::encodeAudioFrame(AudioChunk chunk) {
-	int iArg;
 	float sum;
 	short max;
 
@@ -897,11 +883,10 @@ void AudioInput::encodeAudioFrame(AudioChunk chunk) {
 	QMutexLocker l(&qmSpeex);
 	resetAudioProcessor();
 
-	speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_GET_AGC_GAIN, &iArg);
-	float gainValue = static_cast< float >(iArg);
+	const std::int32_t gainValue = m_preprocessor.getAGCGain();
+
 	if (noiseCancel == Settings::NoiseCancelSpeex || noiseCancel == Settings::NoiseCancelBoth) {
-		iArg = Global::get().s.iSpeexNoiseCancelStrength - iArg;
-		speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_NOISE_SUPPRESS, &iArg);
+		m_preprocessor.setNoiseSuppress(Global::get().s.iSpeexNoiseCancelStrength - gainValue);
 	}
 
 	short psClean[iFrameSize];
@@ -924,7 +909,7 @@ void AudioInput::encodeAudioFrame(AudioChunk chunk) {
 	}
 #endif
 
-	speex_preprocess_run(sppPreprocess, psSource);
+	m_preprocessor.run(*psSource);
 
 	sum = 1.0f;
 	for (unsigned int i = 0; i < iFrameSize; i++)
@@ -942,12 +927,10 @@ void AudioInput::encodeAudioFrame(AudioChunk chunk) {
 						   static_cast< std::streamsize >(iFrameSize * sizeof(short)));
 	}
 
-	spx_int32_t prob = 0;
-	speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_GET_PROB, &prob);
-	fSpeechProb = static_cast< float >(prob) / 100.0f;
+	fSpeechProb = static_cast< float >(m_preprocessor.getSpeechProb()) / 100.0f;
 
 	// clean microphone level: peak of filtered signal attenuated by AGC gain
-	dPeakCleanMic = qMax(dPeakSignal - gainValue, -96.0f);
+	dPeakCleanMic = qMax(dPeakSignal - static_cast< float >(gainValue), -96.0f);
 	float level   = (Global::get().s.vsVAD == Settings::SignalToNoise) ? fSpeechProb : (1.0f + dPeakCleanMic / 96.0f);
 
 	bool bIsSpeech = false;
@@ -1075,12 +1058,10 @@ void AudioInput::encodeAudioFrame(AudioChunk chunk) {
 			}
 		}
 
-		spx_int32_t increment = 0;
-		speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC_INCREMENT, &increment);
+		m_preprocessor.setAGCIncrement(0);
 		return;
 	} else {
-		spx_int32_t increment = 12;
-		speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC_INCREMENT, &increment);
+		m_preprocessor.setAGCIncrement(12);
 	}
 
 	if (bIsSpeech && !bPreviousVoice) {

src/mumble/AudioInput.h

@@ -21,11 +21,11 @@
 #include <vector>
 
 #include <speex/speex_echo.h>
-#include <speex/speex_preprocess.h>
 #include <speex/speex_resampler.h>
 
 #include "Audio.h"
 #include "AudioOutputToken.h"
+#include "AudioPreprocessor.h"
 #include "EchoCancelOption.h"
 #include "MumbleProtocol.h"
 #include "Settings.h"
@@ -224,7 +224,7 @@ class AudioInput : public QThread {
 	static const int iFrameSize = SAMPLE_RATE / 100;
 
 	QMutex qmSpeex;
-	SpeexPreprocessState *sppPreprocess;
+	AudioPreprocessor m_preprocessor;
 	SpeexEchoState *sesEcho;
 
 	/// bResetEncoder is a flag that notifies

src/mumble/AudioStats.cpp

@@ -216,34 +216,28 @@ void AudioNoiseWidget::paintEvent(QPaintEvent *) {
 	paint.fillRect(rect(), pal.color(QPalette::Window));
 
 	AudioInputPtr ai = Global::get().ai;
-	if (!ai.get() || !ai->sppPreprocess)
+	if (!ai.get() || !ai->m_preprocessor)
 		return;
 
 	QPolygonF poly;
 
 	ai->qmSpeex.lock();
 
-	spx_int32_t ps_size = 0;
-	speex_preprocess_ctl(ai->sppPreprocess, SPEEX_PREPROCESS_GET_PSD_SIZE, &ps_size);
-
-	static std::vector< spx_int32_t > noise;
-	noise.resize(static_cast< std::size_t >(ps_size));
-	static std::vector< spx_int32_t > ps;
-	ps.resize(static_cast< std::size_t >(ps_size));
-
-	speex_preprocess_ctl(ai->sppPreprocess, SPEEX_PREPROCESS_GET_PSD, ps.data());
-	speex_preprocess_ctl(ai->sppPreprocess, SPEEX_PREPROCESS_GET_NOISE_PSD, noise.data());
+	const AudioPreprocessor::psd_t ps    = ai->m_preprocessor.getPSD();
+	const AudioPreprocessor::psd_t noise = ai->m_preprocessor.getNoisePSD();
 
 	ai->qmSpeex.unlock();
 
+	assert(ps.size() == noise.size());
+
 	qreal sx, sy;
 
-	sx = (static_cast< float >(width()) - 1.0f) / static_cast< float >(ps_size);
+	sx = (static_cast< float >(width()) - 1.0f) / static_cast< float >(ps.size());
 	sy = static_cast< float >(height()) - 1.0f;
 
 	poly << QPointF(0.0f, height() - 1);
 	float fftmul = 1.0 / (32768.0);
-	for (unsigned int i = 0; i < static_cast< unsigned int >(ps_size); i++) {
+	for (unsigned int i = 0; i < static_cast< unsigned int >(noise.size()); i++) {
 		qreal xp, yp;
 		xp = i * sx;
 		yp = sqrtf(sqrtf(static_cast< float >(noise[i]))) - 1.0f;
@@ -262,7 +256,7 @@ void AudioNoiseWidget::paintEvent(QPaintEvent *) {
 
 	poly.clear();
 
-	for (unsigned int i = 0; i < static_cast< unsigned int >(ps_size); i++) {
+	for (unsigned int i = 0; i < static_cast< unsigned int >(ps.size()); i++) {
 		qreal xp, yp;
 		xp = i * sx;
 		yp = sqrtf(sqrtf(static_cast< float >(ps[i]))) - 1.0f;
@@ -316,7 +310,7 @@ AudioStats::~AudioStats() {
 void AudioStats::on_Tick_timeout() {
 	AudioInputPtr ai = Global::get().ai;
 
-	if (!ai.get() || !ai->sppPreprocess)
+	if (!ai.get() || !ai->m_preprocessor)
 		return;
 
 	bool nTalking = ai->isTransmitting();
@@ -332,22 +326,16 @@ void AudioStats::on_Tick_timeout() {
 	FORMAT_TO_TXT("%06.2f dB", ai->dPeakSignal);
 	qlSignalLevel->setText(txt);
 
-	spx_int32_t ps_size = 0;
-	speex_preprocess_ctl(ai->sppPreprocess, SPEEX_PREPROCESS_GET_PSD_SIZE, &ps_size);
-
-	static std::vector< spx_int32_t > noise;
-	noise.resize(static_cast< std::size_t >(ps_size));
-	static std::vector< spx_int32_t > ps;
-	ps.resize(static_cast< std::size_t >(ps_size));
+	const AudioPreprocessor::psd_t ps    = ai->m_preprocessor.getPSD();
+	const AudioPreprocessor::psd_t noise = ai->m_preprocessor.getNoisePSD();
 
-	speex_preprocess_ctl(ai->sppPreprocess, SPEEX_PREPROCESS_GET_PSD, ps.data());
-	speex_preprocess_ctl(ai->sppPreprocess, SPEEX_PREPROCESS_GET_NOISE_PSD, noise.data());
+	assert(ps.size() == noise.size());
 
 	float s = 0.0f;
 	float n = 0.0001f;
 
-	unsigned int start = static_cast< unsigned int >(ps_size * 300) / SAMPLE_RATE;
-	unsigned int stop  = static_cast< unsigned int >(ps_size * 2000) / SAMPLE_RATE;
+	unsigned int start = static_cast< unsigned int >(ps.size() * 300) / SAMPLE_RATE;
+	unsigned int stop  = static_cast< unsigned int >(ps.size() * 2000) / SAMPLE_RATE;
 
 	for (unsigned int i = start; i < stop; i++) {
 		s += sqrtf(static_cast< float >(ps[i]));
@@ -357,9 +345,7 @@ void AudioStats::on_Tick_timeout() {
 	FORMAT_TO_TXT("%06.3f", s / n);
 	qlMicSNR->setText(txt);
 
-	spx_int32_t v;
-	speex_preprocess_ctl(ai->sppPreprocess, SPEEX_PREPROCESS_GET_AGC_GAIN, &v);
-	float fv = powf(10.0f, (static_cast< float >(v) / 20.0f));
+	float fv = powf(10.0f, (static_cast< float >(ai->m_preprocessor.getAGCGain()) / 20.0f));
 	FORMAT_TO_TXT("%03.0f%%", 100.0f / fv);
 	qlMicVolume->setText(txt);