-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathad_features_extractor.py
204 lines (162 loc) · 7.73 KB
/
ad_features_extractor.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
# -*- coding: utf-8 -*-
"""
Created on Thu Mar 19 01:20:16 2015
@author: xavilizarraga
This function was implemented to extract some features recursively
from a dataset. It is inspired in the streaming_extractor.py from Essentia Library.
So, extract_batch.py will call this function to compute the descriptors.
"""
#! /usr/bin/env python
import sys, os
from os.path import join
import numpy
import pylab
import essentia
import essentia.standard as standard
import essentia.streaming as streaming
from essentia import Pool, INFO
from metadata import readMetadata, getAnalysisMetadata
# maybe I can do the same with my script, create another function like replaygain and call it in the
# process of this function. It will be a good alternative
#import replaygain
#import lowlevel
#import midlevel
#import highlevel
#import panning
#import segmentation
essentia_usage = "usage: \'essentia_extractor [options] config_file input_soundfile output_results\'"
# global defines:
analysisSampleRate = 44100.0
def parse_args():
essentia_version = '%s\n'\
'python version: %s\n'\
'numpy version: %s' % (essentia.__version__, # full version
sys.version.split()[0], # python major version
numpy.__version__) # numpy version
from optparse import OptionParser
parser = OptionParser(usage=essentia_usage, version=essentia_version)
parser.add_option("-v","--verbose",
action="store_true", dest="verbose", default=False,
help="verbose mode")
parser.add_option("-s","--segmentation",
action="store_true", dest="segmentation", default=False,
help="do segmentation")
parser.add_option("-p","--profile",
action="store", type="string", dest="profile", default="music",
help="computation mode: 'music', 'sfx' or 'broadcast'")
parser.add_option("--start",
action="store", dest="startTime", default="0.0",
help="time in seconds from which the audio is computed")
parser.add_option("--end",
action="store", dest="endTime", default="1.0e6",
help="time in seconds till which the audio is computed, 'end' means no time limit")
parser.add_option("--svmpath",
action="store", dest="svmpath", default=join('..', 'svm_models'),
help="path to svm models")
(options, args) = parser.parse_args()
return options, args
def computeAggregation(pool, segments_namespace=''):
stats = ['mean','median','var','min','max','dmean','dmean2','dvar','dvar2' ]
exceptions={'lowlevel.mfcc' : ['mean', 'cov', 'icov']}
for namespace in segments_namespace:
exceptions[namespace+'.lowlevel.mfcc']=['mean', 'cov', 'icov']
if segments_namespace: exceptions['segmentation.timestamps']=['copy']
return standard.PoolAggregator(defaultStats=stats,
exceptions=exceptions)(pool)
def addSVMDescriptors(pool, pathToSvmModels):
#svmModels = [] # leave this empty if you don't have any SVM models
svmModels = ['BAL', 'CUL', 'GDO', 'GRO', 'GTZ', 'PS', 'VI','MAC', 'MAG', 'MEL', 'MHA', 'MPA', 'MRE', 'MSA']
for model in svmModels:
modelFilename = join(pathToSvmModels, model+'.model')
svm = standard.SvmClassifier(model=modelFilename)(pool)
pool.merge(svm)
def computeADLowLevel(input_file, pool, startTime, endTime, namespace=''):
#sampleRate, downmix = getAnalysisMetadata(pool)
sampleRate = 44100
loader = standard.MonoLoader(filename = input_file,sampleRate = sampleRate)
audio = loader()
# Normalize audio data, because vinyl versions were recorded from a mixer
audio = audio/max(audio)
#lowlevel.compute(loader.audio, loader.audio, pool, startTime, endTime, namespace)
# Initialize Essentia Objects
w = standard.Windowing(type = 'hann')
spectrum = standard.Spectrum()
#pool = Pool()
logattacktime = standard.LogAttackTime()
#centroid = Centroid()
hfc = standard.HFC()
energy = standard.Energy()
lowp = standard.LowPass(cutoffFrequency = 80) # 100 Hz - Low pass filter
crest = standard.Crest()
#envelope = streaming.Envelope(attackTime=0.003/5,releaseTime=200./5)
#flatnessDB = streaming.FlatnessDB()
rolloff = standard.RollOff()
#distributionshape = streaming.DistributionShape()
#zerocrossing = streaming.ZeroCrossingRate()
highp = standard.HighPass(cutoffFrequency = 15000) # 100 hz
#centralmoments = streaming.CentralMoments()
strongdecay = standard.StrongDecay()
flatnesssfx = standard.FlatnessSFX()
derivative = standard.Derivative()
entropy = standard.Entropy()
dcremoval = standard.DCRemoval(cutoffFrequency=20)
energybandratiohf = standard.EnergyBandRatio(sampleRate = sampleRate, startFrequency= 10000, stopFrequency = 22050)
energybandratiolf = standard.EnergyBandRatio(sampleRate = sampleRate, startFrequency= 0, stopFrequency = 80)
# Bag Of Features for cd version
for frame in standard.FrameGenerator(audio, frameSize = 2048, hopSize = 128):
frame = dcremoval(frame)
pool.add('lowlevel.logattacktime', logattacktime(frame))
pool.add('lowlevel.lfenergy', energy(lowp(w(frame))))
pool.add('lowlevel.hfenergy', energy(highp(w(frame))))
pool.add('lowlevel.crestfactor',crest(abs(w(frame))))
pool.add('lowlevel.rolloff', rolloff((w(frame))))
pool.add('lowlevel.derivative', numpy.std(abs((derivative(w(frame))))-numpy.median(abs(derivative(w(frame))))))
pool.add('lowlevel.spec_logattacktime', logattacktime(spectrum(w(frame))))
#pool.add('lowlevelcd.centroid', (sampleRate/2.)*(centroid(spectrum(derivative(w(frame))))))
pool.add('lowlevel.hfc', hfc(spectrum(w(frame))))
pool.add('lowlevel.spectralcrest', crest(spectrum(w(frame))))
pool.add('lowlevel.strongdecay', strongdecay(spectrum(w(frame))))
pool.add('lowlevel.flatnesssfx', flatnesssfx(spectrum(w(frame))))
pool.add('lowlevel.hfbandratio', energybandratiohf(spectrum(w(frame))))
pool.add('lowlevel.lfbandratio', energybandratiolf(spectrum(w(frame))))
pool.add('lowlevel.entropy', entropy(spectrum(w(frame))))
"""
def computeMidLevel(input_file, pool, startTime, endTime, namespace=''):
rgain, sampleRate, downmix = getAnalysisMetadata(pool)
loader = streaming.EqloudLoader(filename = input_file,
sampleRate = sampleRate,
startTime = startTime,
endTime = endTime,
replayGain = rgain,
downmix = downmix)
midlevel.compute(loader.audio, pool, startTime, endTime, namespace)
essentia.run(loader)
"""
if __name__ == '__main__':
opt, args = parse_args()
if len(args) != 2: #3:
print "Incorrect number of arguments\n", essentia_usage
sys.exit(1)
#profile = args[0]
input_file = args[0]
output_file = args[1]
pool = Pool()
startTime = float(opt.startTime)
endTime = float(opt.endTime)
# compute descriptors
#readMetadata(input_file, pool)
#INFO('Process step 1: Replay Gain')
#replaygain.compute(input_file, pool, startTime, endTime)
segments_namespace=[]
INFO('Process step 1: Low Level Features')
computeADLowLevel(input_file, pool, startTime, endTime)
#INFO('Process step 2: Mid Level')
#computeMidLevel(input_file, pool, startTime, endTime)
#INFO('Process step 3: High Level')
#highlevel.compute(pool)
# compute statistics
INFO('Process step 4: Aggregation')
stats = computeAggregation(pool, segments_namespace)
# output results to file
INFO('writing results to ' + output_file)
standard.YamlOutput(filename=output_file)(stats)