Merge pull request #1 from Lauchmelder23/development

Development

src/FFT.hpp

@@ -49,7 +49,12 @@ std::vector<std::complex<double>> radix2dit(const std::vector<double>::const_ite
 	return output;
 }
 
-std::vector<std::pair<double, double>> FFT(const std::vector<double>::const_iterator& begin, const std::vector<double>::const_iterator& end, size_t sampleRate)
+std::vector<std::pair<double, double>> 
+FFT(const std::vector<double>::const_iterator& begin, 
+	const std::vector<double>::const_iterator& end, 
+	size_t sampleRate,
+	double minFreq, double maxFreq,
+	unsigned int zeropadding)
 {
 	std::vector<double> signal(begin, end);
 	size_t N = signal.size();
@@ -60,13 +65,21 @@ std::vector<std::pair<double, double>> FFT(const std::vector<double>::const_iter
 		N++;
 	}
 
+	if (zeropadding > 1) {
+		N = (signal.size() << (zeropadding - 1));
+		signal.insert(signal.end(), N - signal.size(), 0);
+	}
+
 	std::vector<std::complex<double>> spectrum = radix2dit(signal.cbegin(), N, 1);
 	double freqRes = (double)sampleRate / (double)N;
 	double nyquistLimit = (double)sampleRate / 2.0f;
 
 	std::vector<std::pair<double, double>> output;
-	double freq = 0.0f;
-	for (int k = 0; freq < nyquistLimit; k++) 
+	double freq = minFreq;
+	if (maxFreq == 0)
+		maxFreq = nyquistLimit;
+
+	for (int k = freq / freqRes; freq < nyquistLimit && freq < maxFreq; k++) 
 	{
 		output.push_back(std::make_pair(freq, 2.0f * std::abs(spectrum[k]) / (double)N));
 		freq += freqRes;

src/main.cpp

@@ -14,6 +14,9 @@ struct Settings {
 	std::vector<std::filesystem::path> files;
 	bool quiet;
 	float splitInterval;
+	double minFreq, maxFreq;
+	unsigned int analyzeChannel;
+	unsigned int zeropadding;
 };
 
 Settings Parse(int argc, char** argv);
@@ -38,6 +41,12 @@ int main(int argc, char** argv)
 		int numChannels = audioFile.getNumChannels();
 
 		nlohmann::json output;
+		int c = setts.analyzeChannel;
+		if (c == 0)
+			c = 1;
+		else
+			numChannels = c;
+
 		for (int c = 1; c <= numChannels; c++) {
 			PRINTER(setts, "\rAnalyzing " << filename << "... Channel " << c << "/" << numChannels << " 0%                  ");
 
@@ -46,7 +55,14 @@ int main(int argc, char** argv)
 
 			if (setts.splitInterval == 0.0f)
 			{
-				std::vector<std::pair<double, double>> spectrum = FFT(audioFile.samples[c-1].cbegin(), audioFile.samples[c-1].cend(), sampleRate);
+				std::vector<std::pair<double, double>> spectrum = 
+					FFT(
+						audioFile.samples[c-1].cbegin(), 
+						audioFile.samples[c-1].cend(), 
+						sampleRate,
+						setts.minFreq, setts.maxFreq,
+						setts.zeropadding
+					);
 
 				output[chName] = nlohmann::json::array();
 				for (const std::pair<double, double>& pair : spectrum) {
@@ -64,8 +80,11 @@ int main(int argc, char** argv)
 							audioFile.samples[c-1].cbegin() + currentSample, 
 							std::min(
 								audioFile.samples[c-1].cbegin() + currentSample + sampleInterval, 
-								audioFile.samples[c-1].cend()), 
-							sampleRate
+								audioFile.samples[c-1].cend()
+							), 
+							sampleRate,
+							setts.minFreq, setts.maxFreq,
+							setts.zeropadding
 						);
 
 					output[chName].push_back({
@@ -102,10 +121,13 @@ Settings Parse(int argc, char** argv)
 		cxxopts::Options options("spectralyze", "Fourier transforms audio files");
 		options
 			.set_width(70)
-			.positional_help("file1 [file2...]")
+			.positional_help("FILE1 [FILE2...]")
 			.add_options()
 			("q,quiet", "Suppress text output", cxxopts::value<bool>()->default_value("false"))
 			("i,interval", "Splits audio file into intervals of length i milliseconds and transforms them individually (0 to not split file)", cxxopts::value<float>())
+			("f,frequency", "Defines the frequency range of the output spectrum (Default: all the frequencies)", cxxopts::value<std::vector<double>>())
+			("p,pad", "Add extra zero-padding. By default, the program will pad the signals with 0s until the number of samples is a power of 2 (this would be equivalent to -p 1). With this option you can tell the program to instead pad until the power of 2 after the next one (-p 2) etc. This increases frequency resolution", cxxopts::value<unsigned int>())
+			("m,mono", "Analyze only the given channel", cxxopts::value<unsigned int>()->default_value("0"))
 			("files", "Files to fourier transform", cxxopts::value<std::vector<std::filesystem::path>>())
 			("h,help", "Print usage")
 			;
@@ -119,6 +141,17 @@ Settings Parse(int argc, char** argv)
 			exit(0);
 		}
 
+		if (!result.count("frequency"))
+		{
+			setts.minFreq = 0.0f;
+			setts.maxFreq = 0.0f;
+		}
+		else
+		{
+			setts.minFreq = result["frequency"].as<std::vector<double>>()[0];
+			setts.maxFreq = result["frequency"].as<std::vector<double>>()[1];
+		}
+
 		if (!result.count("files"))
 		{
 			std::cerr << "At least one positional argument is required." << std::endl;
@@ -128,6 +161,15 @@ Settings Parse(int argc, char** argv)
 		setts.files = result["files"].as<std::vector<std::filesystem::path>>();
 		setts.quiet = (result.count("quiet") ? result["quiet"].as<bool>() : false);
 		setts.splitInterval = (result.count("interval") ? result["interval"].as<float>() : 0.0f);
+		setts.analyzeChannel = (result.count("mono") ? result["mono"].as<unsigned int>() : 0);
+		setts.zeropadding = (result.count("pad") ? result["pad"].as<unsigned int>() : 1);
+		
+
+		if (setts.maxFreq <= setts.minFreq && (setts.maxFreq != 0))
+		{
+			std::cerr << "Maximum frequency cannot be smaller than minimum frequency" << std::endl;
+			exit(1);
+		}
 	}
 	catch (const cxxopts::OptionException& e)
 	{