forked from VladLimbean/thesis_project_audio_recognition
-
Notifications
You must be signed in to change notification settings - Fork 0
/
experiments.py
316 lines (242 loc) · 8.58 KB
/
experiments.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
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
# This script was initially used to test the present implementation
# It is here for reference and it may be integrated a bit later
# However, as of the time of writing the script is not functional
from matplotlib import pyplot as plt
RESULT_DICT = {
'TP': 0,
'TN': 0,
'FP': 0,
'FN': 0,
'FA': 0, # false accept
'FAM_HIT': 0
}
num_tracks, db_tracks = fw.get_wavs_by_fgp(1)
print('Welcome to the experiment script.\nNumber of unique database entries: {}'.format(num_tracks))
test_tracks = ['01.wav',
'aic.wav',
'birds_outside_002_wide.wav',
'busy_dining_room_002.wav',
'madrid_food_market_busy_001.wav',
'rain_umbrella_001_wide.wav',
'50s_elevator_ride_down_001.wav',
'elevator_mechanism_002_wide.wav',
'elevator_mechanism_005_wide.wav',
'soviet_elevator_door_close_001.wav']
mp3_test_tracks = [
'01. Hells Bells.mp3',
'02. Shoot To Thrill.mp3',
'03. What Do You Do For Money Honey.mp3',
'05. Let Me Put My Love Into You.mp3',
'06. Back In Black.mp3',
'07. You Shook Me All Night Long.mp3',
'08. Have A Drink On Me.mp3',
'09. Shake A Leg.mp3',
'01. Hard As A Rock.mp3',
'02. Cover You In Oil.mp3',
'04. Givin The Dog A Bone.mp3',
'10. Rock And Roll Aint Noise Pollution.mp3',
]
def reset_result_dict():
for key in RESULT_DICT.keys():
RESULT_DICT[key] = 0
def exp_for_sensitivity(track_list, dir_map, limit=None):
for tr in track_list:
directory = dir_map[tr]
path = directory + '\\' + tr
sn, list_hash = fw.fingerprint_worker(path, limit=limit)
matches = fw.db.get_matches(list_hash)
result_track, matched_fam, res = fw.align_matches_weighted(matches)
# result track name
r_t_name = result_track['song name']
# check if song is in the DB
db = False
if tr in db_tracks:
db = True
# TP
if r_t_name == tr and db:
RESULT_DICT['TP'] += 1
# TN
elif r_t_name == 'No results found' and not db:
RESULT_DICT['TN'] += 1
# FN
elif r_t_name == 'No results found' and db:
RESULT_DICT['FN'] += 1
# FP
elif r_t_name != tr and db:
RESULT_DICT['FP'] += 1
# FA
elif r_t_name != tr and not db:
RESULT_DICT['FA'] += 1
fam_hit = False
for k, v in matched_fam.items():
if tr in v:
# print('hit! ', song, v)
fam_hit = True
if fam_hit:
RESULT_DICT['FAM_HIT'] += 1
print('Querying {} --- {} s\nResult={}'.format(tr, limit, result_track))
return RESULT_DICT
def exp_with_weighted_align(song, limit=None):
"""Runs a query experiment
Attributes:
song - tuple (is_fgp, song name)
Return:
Dictionary of results
"""
dir_map = export.build_dir_map(export.mpeg_root)
song_in = song[1]
if song_in in dir_map:
directory = dir_map[song_in]
else:
print(song_in, ' missing')
return
sn, list_hash = fw.fingerprint_worker(directory + '\\' + song_in, limit=limit)
matches = fw.db.get_matches(list_hash)
result_track, matched_fam, res = fw.align_matches_weighted(matches)
# result track name
r_t_name = result_track['song name']
# check if song is in the DB
db = False
if song_in in db_tracks:
db = True
# TP
if r_t_name == song_in and db:
RESULT_DICT['TP'] += 1
# TN
elif r_t_name == 'No results found' and not db:
RESULT_DICT['TN'] += 1
# FN
elif r_t_name == 'No results found' and db:
RESULT_DICT['FN'] += 1
# FP
elif r_t_name != song_in and db:
RESULT_DICT['FP'] += 1
# FA
elif r_t_name != song_in and not db:
RESULT_DICT['FA'] += 1
fam_hit = False
for k, v in matched_fam.items():
if song_in in v:
# print('hit! ', song, v)
fam_hit = True
if fam_hit:
RESULT_DICT['FAM_HIT'] += 1
print('Querying {} --- {} s\nResult={}'.format(song_in, limit, result_track))
return RESULT_DICT
def run_sensitivity_test():
limits = [1, 2, 4, 8]
list_tracks, dir_map = generate_track_list('wav', db_set=40, hold_back_set=10)
for lim in limits:
reset_result_dict()
result = exp_for_sensitivity(list_tracks, dir_map, limit=lim)
print('Limit: {} s'.format(lim))
print(result)
def run_exp4_align_weighted():
limits = [1, 2, 4, 8]
#limits = [4]
#num_tracks, track = fw.get_wavs_by_fgp(1)
result = None
for l in limits:
# reset the result dictionary for different limits
reset_result_dict()
counter = 0
for t in db_tracks:
if counter == 500:
break
counter += 1
t = [1, t]
result = exp_with_weighted_align(t, l)
print('Limit: {} s'.format(l))
print(result)
def generate_track_list(track_type, db_set=0, hold_back_set=0):
"""Generates a list of tracks containing a specified number of
tracks from the database and specified number of tracks not present
in the database.
This is done for the purpose of specificity testing.
Attributes:
db_tracks - number of tracks we want from the database
hold_back - number of tracks we want that are not in the database
track_type - wav or mp3
Return:
res - a list of audio track titles
"""
# counters for how many songs we've added for each set
goal_db = 0
goal_hold_back = 0
res = []
# get all available tracks
if track_type == 'wav':
dir_map = export.build_dir_map(export.wav_root)
elif track_type == 'mp3':
dir_map = export.build_dir_map(export.mpeg_root)
else:
print('Only wav or mp3 types accepted')
return
# get number of tracks and a list of tracks
nt, all_tracks_from_db = fw.get_wavs_by_fgp(1)
# create the list
for tr in dir_map.keys():
if tr in all_tracks_from_db and goal_db < db_set:
res.append(tr)
goal_db += 1
if tr not in all_tracks_from_db and goal_hold_back < hold_back_set:
res.append(tr)
goal_hold_back += 1
if goal_db == db_set and goal_hold_back == hold_back_set:
print('done!')
break
return res, dir_map
def exp_aligned_matches():
song_path = export.db_test + '\\' + 'c1.wav'
sn, list_hashes = fw.fingerprint_worker(song_path, limit=4)
print('Querying: {}'.format(sn))
matches = fw.db.get_matches(list_hashes)
song, fam_dict, cand = fw.align_matches_weighted(matches)
print('Most likely res = {}\n'.format(song))
sim.compute_sim(song['song name'], cand)
def test_list_hash_colision_rate(test_track, limit=None):
name, list_hash = fw.fingerprint_worker(test_track, limit)
return list_hash
def run_test_list_colision_rate():
test_track = 'D:\\thesis-data\\c2.wav'
_, list_hash_all = fw.fingerprint_worker(test_track)
print('len=', len(list_hash_all))
l_num_hash = []
num_col_hash = []
for l in range(0, 42):
print('limit= ', l)
limited_list = test_list_hash_colision_rate(test_track, limit=l)
num_hash = len(limited_list)
intersect = len(limited_list.intersection(list_hash_all))
l_num_hash.append(num_hash)
num_col_hash.append(intersect)
print('Num hash: {}'.format(num_hash))
print('intersection=', intersect)
plt.plot(l_num_hash, num_col_hash)
plt.show()
print('num hash',l_num_hash)
print('num intersect', num_col_hash)
return l_num_hash, num_col_hash
def test_all_answers(song_in):
"""Tests whether algorithm correctly returns TP, TN, FP, FN"""
dir_map = export.build_dir_map(export.wav_root)
if song_in in dir_map:
directory = dir_map[song_in]
else:
print(song_in, ' missing')
return
sn, list_hash = fw.fingerprint_worker(directory + '\\' + song_in, limit=1)
matches = fw.db.get_matches(list_hash)
result_track, matched_fam, res = fw.align_matches_weighted(matches)
for e in res:
print(e)
# result track name
r_t_name = result_track['song name']
print(r_t_name)
if __name__ == '__main__':
path = '' # include the path to your song
limit = 0 # how many seconds to check from the song
sn, list_hash = fw.fingerprint_worker(path, limit=limit)
matches = fw.db.get_matches(list_hash)
result_track, matched_fam, res = fw.align_matches_weighted(matches)
print('Recognized track=',result_track)