e@0: # -*- coding: utf-8 -*-
e@0: """
e@0: Created on Mon Jun  1 11:42:06 2015
e@0: 
e@0: @author: Emmanouil Theofanis Chourdakis
e@0: """
e@0: 
e@0: # Note, reference everything!
e@0: 
e@0: from sys import argv
e@0: 
e@0: 
e@0: if __name__=="__main__":
e@0:     if len(argv) != 3:
e@0:         print("Incorrect number of arguments:")
e@0:         print("Usage: ")
e@0:         print("%s <input> <output>")
e@0:         print("")
e@0:         print("Arguments:")
e@0:         print("<input>\tThe input filename. Can be .wav, .mp3, etc...")
e@0:         print("<parameters>\t The parameters filename, in .yaml format" )
e@0:         print("<output>\tThe output filename in .yaml format")        
e@0:         sys.exit(-1)
e@0:     else:
e@0:         
e@0:         print("[II] Loading libraries")
e@0:         
e@0:         import essentia
e@0:         from essentia import Pool
e@0:         from essentia.standard import  *
e@0:         import yaml
e@0:         
e@0:         
e@0:         # reqyures matplotlib
e@0:         from pylab import *
e@0:         
e@0:         #requires numpy
e@0:         from numpy import *
e@0:         
e@0:         #requires scikit-learn
e@0:         from sklearn.metrics import pairwise_distances        
e@0:         
e@0:         d = {}
e@0:         v = {}
e@0:         
e@0:         fname = argv[1]
e@0:         outfname = argv[2]
e@0:         
e@0: 
e@0: 
e@0:         trackname = fname.split('.')[0].split('/')[-1]  
e@0: 
e@0:         
e@0:         if outfname.partition('.')[-1].lower() not in ['json', 'yaml']:
e@0:             print("Please choose a .json or .yaml as an output file.")
e@0:             sys.exit(-1)
e@0:         else:
e@0:             if outfname.partition('.')[-1].lower() == 'json':
e@0:                 output = YamlOutput(filename = outfname, format='json') 
e@0:             else:
e@0:                 output = YamlOutput(filename = outfname, format='yaml')
e@0:         
e@0:         print("Feature extraction of `%s\'" % fname)
e@0: 
e@0:         # Sampling Rate
e@0:         SR = 16000.0
e@0:         
e@0:         # Sampling Frequency
e@0:         T = 1.0/SR
e@0: 
e@0:         # FrameSize
e@0:         tframeSize = 23 #ms    
e@0:         frameSize = int(ceil(tframeSize*SR/1000)) if mod(ceil(tframeSize*SR/1000),2) == 0 \
e@0:                                                   else int(floor(tframeSize*SR/1000))
e@0:                                                   
e@0:         # HopSize
e@0:         hopSize = frameSize/2
e@0:         
e@0:         # Load Audio
e@0:         audio = MonoLoader(filename = fname, sampleRate=SR)()
e@0:         
e@0:         
e@0:         #Window Frames
e@0:         w = Windowing(size = frameSize, type = 'hamming')    
e@0:         
e@0:         # Spectrum
e@0:         spec = Spectrum(size=1024)
e@0:         
e@0:         # Pool to append mean and variance
e@0:         pool = Pool()
e@0:         globalPool = Pool()
e@0:         
e@0:         # Below are Features to be used in the feature extraction stage
e@0:         # We use, Spectral Contrast, MFCCs, Zero-Crossing rate, RMS, 
e@0:         # Crest Factor, Spectral Centroid, Spectral Occupation, Spectral Flux
e@0:         
e@0:         # Spectral Contrast
e@0:         sc =  SpectralContrast(frameSize = frameSize, highFrequencyBound = 8000, sampleRate = SR)        
e@0:         
e@0:         # MFCCs
e@0:         mfccs = MFCC(highFrequencyBound = 8000, sampleRate = SR)
e@0: 
e@0:         # Spectral Centroid
e@0:         centroid = Centroid(range = SR/2)
e@0:         
e@0:         # Spectral Roll-Off
e@0:         rolloff = RollOff(sampleRate = SR, cutoff = 0.9)
e@0:         
e@0:         # Spectral Flux
e@0:         flux = Flux()
e@0:         
e@0:         # Zero Crossing Rate
e@0:         zcr = ZeroCrossingRate()        
e@0: 
e@0:         # RMS 
e@0:         rms = RMS()
e@0:         
e@0:         # Crest Factor
e@0:         crest = Crest()
e@0: 
e@0:         
e@0: 
e@0:         
e@0:         # Segmentation based on Onset detection-based temporal modeling
e@0:         print("[II] Calculating features for %s, please wait..." % fname)        
e@0:         # Onset Detection
e@0:         
e@0:         print("[II] Splitting to onsets...")
e@0:     
e@0:         onsetdetection = OnsetDetectionGlobal(frameSize = frameSize, hopSize = hopSize, sampleRate = SR)(audio)
e@0:         onsets = Onsets()(essentia.array([onsetdetection]), [1])
e@0: 
e@0:             
e@0:         
e@0:         print("[II] done, extracting features...")
e@0:         for o in range(0, len(onsets)-1): 
e@0:             IOI = audio[onsets[o]*SR:onsets[o+1]*SR]
e@0:             
e@0:             
e@0:     
e@0:             
e@0:             if len(IOI) == 0:
e@0:                 break;
e@0:                 
e@0:             
e@0:             for frame in FrameGenerator(IOI, frameSize, hopSize):
e@0:                 # Temporal Features
e@0:             
e@0:                 zerocrossingrate = zcr(frame)              
e@0:                 rmsvalues = rms(frame)
e@0:                 
e@0:                 # Spectral features
e@0:                 framespectrum = spec(w(frame))
e@0:                 framecontrast = sc(framespectrum)
e@0:                 mfcc_coeffs = mfccs(framespectrum)[1]
e@0:                 spectralcentroid = centroid(framespectrum)
e@0:                 spectralrolloff = rolloff(framespectrum)
e@0:                 spectralflux = rolloff(framespectrum)
e@0:                                 
e@0:                 
e@0:                 
e@0:                 pool.add('lowlevel.zcr', zerocrossingrate)
e@0:                 pool.add('lowlevel.rms', rmsvalues)
e@0:                 pool.add('lowlevel.spectrum.centroid', spectralcentroid)
e@0:                 pool.add('lowlevel.spectrum.rolloff', spectralrolloff)
e@0:                 pool.add('lowlevel.mfcc.coeffs', mfcc_coeffs)
e@0:                 pool.add('lowlevel.spectrum.magnitude', framespectrum)
e@0:                 pool.add('lowlevel.contrast.contrast',  framecontrast[0])  
e@0:                 pool.add('lowlevel.contrast.valleys',  framecontrast[1])            
e@0:                 pool.add('lowlevel.spectrum.flux', spectralflux)
e@0:                 
e@0:                 
e@0:             
e@0:             spectrumfull = pool['lowlevel.spectrum.magnitude']
e@0:             spectralcontrast = pool['lowlevel.contrast.contrast']
e@0:             spectralvalleys = pool['lowlevel.contrast.valleys']
e@0:             spectralcentroidfeature = pool['lowlevel.spectrum.centroid']
e@0:             spectralrollofffeature = pool['lowlevel.spectrum.rolloff']
e@0:             spectralfluxfeature = pool['lowlevel.spectrum.flux']
e@0:             
e@0:             spectralfeature = concatenate((spectralcontrast,spectralvalleys),1)
e@0:             mfccfeature = pool['lowlevel.mfcc.coeffs']
e@0:             zcrfeature = pool['lowlevel.zcr']
e@0:             rmsfeature = pool['lowlevel.rms']
e@0:             crestfeature = crest(rmsfeature)
e@0:             1
e@0:             
e@0:             meanspectralfeature = mean(spectralfeature, 0)
e@0:             for i in range(0, shape(spectralfeature)[1]):
e@0:                 globalPool.add('spectralcontrast_%d' % i , meanspectralfeature[i])
e@0:             globalPool.add('spectralcentroid', mean(spectralcentroidfeature, 0))
e@0:             globalPool.add('spectralrolloff', mean(spectralrollofffeature, 0))
e@0:             globalPool.add('spectralflux', mean(spectralfluxfeature, 0))
e@0:             
e@0:             # Expand mfccs
e@0:             meanmfcc = mean(mfccfeature, 0)
e@0:             for i in range(0, shape(mfccfeature)[1]):
e@0:                 globalPool.add('mfcc_%d' % i, meanmfcc[i])
e@0:                 
e@0:             
e@0:             globalPool.add('zcr', mean(zcrfeature, 0))
e@0:             globalPool.add('rms', mean(rmsfeature, 0))
e@0:             globalPool.add('crest', crestfeature)
e@0:             
e@0:             pool.clear()
e@0: 
e@0:         print("[II] done.")  
e@0:         
e@0:         
e@0:         print("[II] Saving data to %s:" % outfname)
e@0:         globalPool.add("metadata.filename", fname)
e@0:         output(globalPool)        
e@0:         
e@0: