added agc; added normalization to captured samples; some bug fixes

Source/Processor.cpp

@@ -26,6 +26,15 @@ Processor::~Processor()
 {
 }
 
+double Processor::getAgcGained(double sample)
+{
+	double gained = sample * agcGain;
+	double _agcSpeed = agcSpeed / frameSize;
+	if (std::abs(gained) > 1) agcGain -= agcSpeed;
+	else agcGain += _agcSpeed;
+	return gained;
+}
+
 void Processor::addSamples(std::vector<double> samples)
 {
 	for (int i = 0; i < samples.size(); ++i)
@@ -197,6 +206,8 @@ void Processor::loadState(std::string jsonText)
 
 	threshold = jsonObj["threshold"].get<double>();
 
+	reload();
+
 }
 
 double Processor::getThreshold()
@@ -221,12 +232,13 @@ bool Processor::process()
 	for (int i = 0; i < frameSize; ++i)
 	{
 		uint16_t ptr = bufferPtr - frameSize + i;
-		frame[i] = buffer[ptr];
+		frame[i] = getAgcGained(buffer[ptr]);
 	}
 
 	if (capturingSample)
 	{
 		SoundDetector::getSpectrum(frame);
+		SoundDetector::normalize(frame);
 		sampleSpectrum[capturingSampleId] = frame;
 		capturingSample = false;
 		reload();
@@ -239,13 +251,13 @@ bool Processor::process()
 		if (sampleEnabled[i])
 		{
 			double distance = detectors[i].process(frame);
-			mfccScore = std::max((1 - distance - 0.35) * 3, mfccScore);
+			mfccScore = std::max((1 - distance - 0.48) * 10, mfccScore);
 		}
 	}
-	currentMfccScore = std::min(mfccScore, 1.);
+	currentMfccScore = std::max(std::min(mfccScore, 1.), 0.);
 
-	//OutputDebugString(std::to_string(currentMfccScore).c_str());
-	//OutputDebugString("\n");
+	OutputDebugString(std::to_string(currentMfccScore).c_str());
+	OutputDebugString("\n");
 
 	bool result = currentMfccScore > threshold;
 

Source/Processor.h

@@ -42,11 +42,13 @@ class Processor
 	bool capturingSample = false;
 	int capturingSampleId;
 	Processor *getMuteTimeFrom;
+	double getAgcGained(double sample);
 
 	double mfccScoreOffset;
 	double mfccScoreScale;
 	double threshold = 0.5;
-
+	double agcSpeed = 0.2;
+	double agcGain = 1;
 
 };
 

Source/SoundDetector.cpp

@@ -101,6 +101,23 @@ void SoundDetector::getSpectrum(std::vector<double>& frame)
 	}
 }
 
+void SoundDetector::normalize(std::vector<double>& frame)
+{
+	double sum = 0;
+	for (int i = 0; i < frame.size(); ++i)
+	{
+		sum += frame[i];
+	}
+	if (sum != 0)
+	{
+		sum /= frame.size();
+		for (int i = 0; i < frame.size(); ++i)
+		{
+			frame[i] /= sum;
+		}
+	}
+}
+
 void SoundDetector::_process(std::vector<double> frame)
 {
 	getSpectrum(frame);
@@ -122,7 +139,7 @@ void SoundDetector::_process(std::vector<double> frame)
 
 	int scaled_size = std::min((int)(ratio * 300), (int)frame.size() / 2);
 
-	if (ratio > 2 || ratio < 0.5) scaled_size = 0;
+	if (ratio > 2 || ratio < 0.5 || !std::isfinite(ratio)) scaled_size = 0;
 
 	std::vector<double> scaled_sample_spectrum(scaled_size);
 
@@ -144,7 +161,7 @@ void SoundDetector::_process(std::vector<double> frame)
 		sum_ += pow(frame[i], 2) * pow(scaled_sample_spectrum[i], 2) / std::pow(scaled_size, 4);
 	}
 
-	calculatedDistance = distance - sum_ / 5;
+	calculatedDistance = distance - sum_ / 3;
 }
 
 double SoundDetector::process(std::vector<double> frame)

Source/SoundDetector.h

@@ -15,6 +15,7 @@ class SoundDetector
 	~SoundDetector();
 
 	static void getSpectrum(std::vector<double>& frame);
+	static void normalize(std::vector<double>& frame);
 
 private:
 	void _process(std::vector<double> frame);