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@0: 
d@0: AudioReader::AudioReader()
d@0: {
d@0: 	m_formatManager.registerBasicFormats();
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@0: 	std::vector<FeatureData> featureData = std::vector<FeatureData>();
d@0: 	std::string CSVFileName = "..\\FeatureData" + writeCSV.currentDateTime();
d@0: 		
d@0: 	vector<string> audioFileNames = audioFileData.GetFileNames();
d@0: 	vector<string> labels = audioFileData.GetLabels();
d@0: 
d@0: 	//AudioSourceFeatureExtractor audioSourceFeatureExtractor = AudioSourceFeatureExtractor();	
d@0: 	m_AudioSourceFeatureExtractor.Initialise(SAMPLERATE);
d@0: 
d@0: 	for(size_t i=0; i<audioFileNames.size(); i++)
d@0: 	{
d@0: 		float percentcomplete = (float)i / (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@0: 
d@0: 		//Create file from    our audio data
d@0: 		File audioFile(audioFileNames[i].c_str());
d@0: 		
d@0: 		AudioFormatReader* m_audioFileReader = m_formatManager.createReaderFor(audioFile);		
d@0: 
d@0: 		if(m_audioFileReader != NULL)
d@0: 		{
d@0: 			m_fSampleRate		= (float)(m_audioFileReader->sampleRate);
d@0: 			m_iLengthInSamples  = (int)m_audioFileReader->lengthInSamples;
d@0: 			m_iNumOfChannels    = m_audioFileReader->numChannels;
d@0: 
d@0: 			if (m_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@0: 			
d@0: 			//Get loudest 30 secs. of audio
d@0: 			int numOfSamplesToCollect = (int)(analysisWindowSize * m_fSampleRate);
d@0: 
d@0: 			if(m_iLengthInSamples <= numOfSamplesToCollect)
d@0: 			{
d@0: 				numOfSamplesToCollect = m_iLengthInSamples;
d@0: 			}
d@0: 
d@0: 			//Length of the full track in stereo;
d@0: 			int* destSamples[2] = {0};
d@0: 			int* L = new int[(size_t)(m_iLengthInSamples)]; 
d@0: 			memset(L, 0, (size_t)m_iLengthInSamples*sizeof(float)); 
d@0: 			destSamples[0]=L;
d@0: 			destSamples[1]=L;				
d@0: 	
d@0: 			//30 sec clips to check energy levels
d@0: 			float* destSamplesFloat = new float[(size_t)numOfSamplesToCollect];
d@0: 			memset(destSamplesFloat, 0, (size_t)numOfSamplesToCollect*sizeof(float)); 
d@0: 
d@0: 			////30 sec clips to check energy levels
d@0: 			//float* destSamplesFloatLoudest = new float[(size_t)numOfSamplesToCollect];
d@0: 			//memset(destSamplesFloatLoudest, 0, (size_t)numOfSamplesToCollect*sizeof(float)); 
d@0: 
d@0: 			int timesToLoop = 0;
d@0: 
d@0: 			if(m_iLengthInSamples == numOfSamplesToCollect)
d@0: 			{
d@0: 				timesToLoop = 1;
d@0: 			}
d@0: 			else
d@0: 			{
d@0: 				timesToLoop = (int)((m_iLengthInSamples - numOfSamplesToCollect) / (ENERGYSEARCHTIME * m_fSampleRate));
d@0: 			}
d@0: 
d@0: 			std::vector<float> thirtySecEnergy = std::vector<float>();
d@0: 
d@0: 			//float loudestEnergy = 0.0;
d@0: 
d@0: 			m_audioFileReader->readSamples(destSamples, 2, 0, 0, m_iLengthInSamples);
d@0: 
d@0: 			for(int j=0; j < timesToLoop;j++)
d@0: 			{	
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@0: 					if(m_iNumOfChannels > 1)
d@0: 					{
d@0: 						destSamplesFloat[n] = ((float)((destSamples[0][int(j * ENERGYSEARCHTIME * m_fSampleRate) + n] + destSamples[1][int( j * ENERGYSEARCHTIME * m_fSampleRate) + n]) / 2) / (0x7fffffff));
d@0: 					}
d@0: 					else
d@0: 					{
d@0: 						destSamplesFloat[n] = ((float)(destSamples[0][int(j * ENERGYSEARCHTIME * m_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@0: 			//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@0: 			int* L30 = new int[(size_t)numOfSamplesToCollect];
d@0: 			memset(L30, 0, (size_t)numOfSamplesToCollect*sizeof(float));
d@0: 			thirtySecSamples[0]=L30; //Left channel
d@0: 			thirtySecSamples[1]=L30; //Left right		
d@0: 
d@0: 			//Read the 30 secs. in
d@0: 			m_audioFileReader->readSamples(thirtySecSamples, 2, 0, int(maxIdx * ENERGYSEARCHTIME * m_fSampleRate), numOfSamplesToCollect);
d@0: 			memset(destSamplesFloat, 0, (size_t)numOfSamplesToCollect*sizeof(float)); 
d@0: 
d@0: 			for(int n=0; n<numOfSamplesToCollect; n++)
d@0: 			{			
d@0: 				//Sum to mono if needed 
d@0: 				if(m_iNumOfChannels > 1)
d@0: 				{
d@0: 					destSamplesFloat[n] = ((float)((thirtySecSamples[0][n] + thirtySecSamples[1][n]) / 2) / (0x7fffffff));
d@0: 				}
d@0: 				else
d@0: 				{
d@0: 					destSamplesFloat[n] = ((float)(thirtySecSamples[0][n]) / (0x7fffffff));
d@0: 				}
d@0: 			}
d@0: 			
d@0: 			std::vector<ObservationData> newObs =  m_AudioSourceFeatureExtractor.Process(destSamplesFloat, numOfSamplesToCollect);	
d@0: 
d@0: 			FeatureData newFeature = FeatureData(newObs, labels[i], audioFileNames[i], m_fSampleRate, FFTSIZE, (float)numOfSamplesToCollect, poolTimeSecs);
d@0: 		
d@0: 			writeCSV.Write(CSVFileName, newFeature);
d@0: 			
d@0: 			//Update the screen information;
d@0: 			cout << cout << string(outputStr.length(),'\b');		
d@0: 		
d@0: 			//Cleanup
d@0: 			if(L != NULL)
d@0: 			{
d@0: 				delete[] L;
d@0: 				L = nullptr;
d@0: 			}
d@0: 
d@0: 			if(destSamplesFloat != NULL)
d@0: 			{
d@0: 				delete[] destSamplesFloat;
d@0: 				destSamplesFloat = nullptr;
d@0: 				//destSamplesFloatLoudest = nullptr;
d@0: 			}
d@0: 
d@0: 			//if(destSamplesFloatLoudest != NULL)
d@0: 			//{
d@0: 			//	delete[] destSamplesFloatLoudest;
d@0: 			//	
d@0: 			//}
d@0: 		}
d@0: 		else
d@0: 		{
d@0: 			cout << "\n\n\nERROR: Could not find file!!!\n\n\n";
d@0: 		}
d@0: 
d@0: 		//Cleanup
d@0: 		if(m_audioFileReader != NULL)
d@0: 		{
d@0: 			delete[] m_audioFileReader;
d@0: 			m_audioFileReader  = nullptr;
d@0: 		}
d@0: 	}
d@0: 
d@0: 	m_AudioSourceFeatureExtractor.Finalize();
d@0: 	
d@0: 	return featureData;
d@0: };