d@0: /*
d@0:   ==============================================================================
d@0: 
d@0:     AudioReader.cpp
d@0:     Created: 27 Aug 2014 3:17:10pm
d@0:     Author:  david.ronan
d@0: 
d@0:   ==============================================================================
d@0: */
d@0: 
d@0: 
d@0: 
d@0: 
d@0: #include "AudioReader.h"
d@0: #include "WriteCSV.h"
d@0: #include <iostream>
d@0: #include <vector>
d@0: 
d@0: #define ENERGYSEARCHTIME 3
d@0: #define SAMPLERATE 22050.0f
d@13: #include "XMLWrite.h"
d@0: 
d@0: AudioReader::AudioReader()
d@2: {
d@1: 
d@0: };
d@0: 
d@0: AudioReader::~AudioReader()
d@0: {
d@0: 
d@0: };
d@0: 
d@0: std::vector<FeatureData> AudioReader::Read(AudioFileData audioFileData, float poolTimeSecs, int analysisWindowSize)
d@0: {
d@0: 	WriteCSV writeCSV = WriteCSV();
d@13: 	XMLWrite xmlWrite = XMLWrite();
d@0: 	std::vector<FeatureData> featureData = std::vector<FeatureData>();
d@0: 	std::string CSVFileName = "..\\FeatureData" + writeCSV.currentDateTime();
d@1: 
d@0: 	vector<string> audioFileNames = audioFileData.GetFileNames();
d@2: 	vector<string> labels = audioFileData.GetLabels();
d@2: 
d@1: 	AudioFormatManager m_formatManager;
d@1:     m_formatManager.registerBasicFormats();
d@1: 	m_AudioSourceFeatureExtractor.Initialise(SAMPLERATE);
d@0: 
d@0: 	for(size_t i=0; i<audioFileNames.size(); i++)
d@0: 	{
d@3: 		float percentcomplete = static_cast<float>(i) / static_cast<float>(audioFileNames.size()) * 100.0f;
d@0: 
d@0: 		std::string outputStr = "Extracting features for " + audioFileNames[i] + "\n" +  std::to_string(percentcomplete) + "% complete...";
d@0: 		cout << outputStr;
d@9: 	
d@0: 		//Create file from    our audio data
d@0: 		File audioFile(audioFileNames[i].c_str());
d@0: 
d@9: 		unique_ptr<AudioFormatReader> audioFileReader(m_formatManager.createReaderFor(audioFile));
d@1: 
d@3: 		if(audioFileReader != nullptr)
d@3: 		{
d@3: 			float fSampleRate		= static_cast<float>(audioFileReader->sampleRate);
d@3: 			int iLengthInSamples  = static_cast<int>(audioFileReader->lengthInSamples);
d@3: 			int iNumOfChannels    = audioFileReader->numChannels;
d@1: 
d@3: 			if (fSampleRate != 22050.0f)
d@0: 			{
d@0: 				cout << "\n\n\nERROR: File is not the required 22050 Hz sample rate.!!!\n\n\n";
d@0: 			}
d@1: 
d@0: 			//Get loudest 30 secs. of audio
d@3: 			int numOfSamplesToCollect = static_cast<int>(analysisWindowSize * fSampleRate);
d@0: 
d@3: 			if(iLengthInSamples <= numOfSamplesToCollect)
d@0: 			{
d@3: 				numOfSamplesToCollect = iLengthInSamples;
d@0: 			}
d@0: 
d@0: 			//Length of the full track in stereo;
d@9: 			int* destSamples[2] = { nullptr };
d@9: 			unique_ptr<int> L(new int[iLengthInSamples]);
d@9: 			memset(L.get(), 0, iLengthInSamples);
d@9: 			destSamples[0] = L.get();
d@9: 			destSamples[1] = L.get();
d@1: 
d@9: 			
d@0: 			//30 sec clips to check energy levels
d@9: 			std::vector<float> destSamplesFloat = std::vector<float>(static_cast<size_t>(numOfSamplesToCollect, 0));
d@0: 
d@1: 			int timesToLoop = 0;
d@0: 
d@3: 			if(iLengthInSamples == numOfSamplesToCollect)
d@0: 			{
d@0: 				timesToLoop = 1;
d@0: 			}
d@0: 			else
d@0: 			{
d@3: 				timesToLoop = static_cast<int>((iLengthInSamples - numOfSamplesToCollect) / (ENERGYSEARCHTIME * fSampleRate));
d@0: 			}
d@0: 
d@0: 			std::vector<float> thirtySecEnergy = std::vector<float>();
d@0: 
d@0: 			//float loudestEnergy = 0.0;
d@9: 			bool dave;
d@9: 			dave = audioFileReader->readSamples(destSamples, iNumOfChannels, 0, 0, iLengthInSamples);
d@0: 
d@0: 			for(int j=0; j < timesToLoop;j++)
d@1: 			{
d@0: 				float fSum=0.f;
d@0: 
d@0: 				for(int n=0; n<numOfSamplesToCollect; n++)
d@0: 				{
d@0: 					//Sum to mono if needed and workout the energy for each 30 sec. frame
d@3: 					if(iNumOfChannels > 1)
d@0: 					{
d@9: 						destSamplesFloat.push_back(static_cast<float>((destSamples[0][int(j * ENERGYSEARCHTIME * fSampleRate) + n] + destSamples[1][int( j * ENERGYSEARCHTIME * fSampleRate) + n]) / 2) / (0x7fffffff));
d@0: 					}
d@0: 					else
d@0: 					{
d@9: 						destSamplesFloat.push_back(static_cast<float>(destSamples[0][int(j * ENERGYSEARCHTIME * fSampleRate) + n]) / (0x7fffffff));
d@0: 					}
d@0: 
d@0: 					fSum+=(destSamplesFloat[n] * destSamplesFloat[n]);
d@0: 				}
d@0: 
d@0: 				//Normalise and push onto the list of 30 sec clips.
d@0: 				fSum /= numOfSamplesToCollect;
d@0: 
d@0: 				//if (fSum > loudestEnergy)
d@0: 				//{
d@0: 				//	loudestEnergy = fSum;
d@0: 				//	destSamplesFloatLoudest = destSamplesFloat;
d@0: 				//}
d@0: 
d@0: 				thirtySecEnergy.push_back(fSum);
d@0: 			}
d@0: 
d@1: 			//Find the index of the section with the most energy
d@0: 			int maxIdx = std::distance(thirtySecEnergy.begin(), max_element(thirtySecEnergy.begin(), thirtySecEnergy.end()));
d@0: 
d@0: 			int* thirtySecSamples[2] = {0};
d@9: 			unique_ptr<int> L30(new int[iLengthInSamples]{ 0 });
d@9: 			memset(L30.get(), 0, iLengthInSamples);
d@9: 			thirtySecSamples[0]=L30.get(); //Left channel
d@9: 			thirtySecSamples[1]=L30.get(); //Left right
d@0: 
d@0: 			//Read the 30 secs. in
d@3: 			audioFileReader->readSamples(thirtySecSamples, iNumOfChannels, 0, int(maxIdx * ENERGYSEARCHTIME * fSampleRate), numOfSamplesToCollect);
d@9: 
d@9: 			destSamplesFloat = std::vector<float>(static_cast<size_t>(numOfSamplesToCollect, 0));
d@0: 
d@0: 			for(int n=0; n<numOfSamplesToCollect; n++)
d@1: 			{
d@1: 				//Sum to mono if needed
d@3: 				if(iNumOfChannels > 1)
d@0: 				{
d@9: 					destSamplesFloat.push_back(static_cast<float>((thirtySecSamples[0][n] + thirtySecSamples[1][n]) / 2) / (0x7fffffff));
d@0: 				}
d@0: 				else
d@0: 				{
d@9: 					destSamplesFloat.push_back(static_cast<float>(thirtySecSamples[0][n]) / (0x7fffffff));
d@0: 				}
d@0: 			}
d@0: 
d@9: 			std::vector<ObservationData> newObs =  m_AudioSourceFeatureExtractor.Process(destSamplesFloat);
d@1: 
d@3: 			FeatureData newFeature = FeatureData(newObs, labels[i], audioFileNames[i], fSampleRate, FFTSIZE, static_cast<float>(numOfSamplesToCollect), poolTimeSecs);
d@1: 
d@14: 			//XML from now on.
d@0: 			writeCSV.Write(CSVFileName, newFeature);
d@13: 			xmlWrite.Write(CSVFileName, newFeature);
d@1: 
d@0: 			//Update the screen information;
d@2: 			
d@9: 			cout << string(outputStr.length(),'\b');	
d@3: 
d@0: 		}
d@0: 		else
d@0: 		{
d@0: 			cout << "\n\n\nERROR: Could not find file!!!\n\n\n";
d@3: 		}		
d@0: 	}
d@0: 
d@0: 	m_AudioSourceFeatureExtractor.Finalize();
d@1: 
d@0: 	return featureData;
d@2: };