-
Notifications
You must be signed in to change notification settings - Fork 0
/
OpenSMILE.py
37 lines (30 loc) · 1.2 KB
/
OpenSMILE.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
import opensmile
import pandas as pd
import glob
from multiprocessing import Pool, cpu_count
import time
import re
regex = r"\\(\w*)\\"
smile = opensmile.Smile(feature_set=opensmile.FeatureSet.ComParE_2016, feature_level=opensmile.FeatureLevel.Functionals,) # Select the features set
path = "Sounds"
data = []
files = glob.glob(path + '/**/*.wav', recursive= True) # Get all the wav files in the subdirectories
def process(file): # Extract the features with OpenSMILE
try:
df = smile.process_file(file)
df["targetClass"] = re.findall(regex, file, re.MULTILINE)[0]
return df
except Exception as inst:
print(type(inst))
return []
if __name__ == "__main__":
print("Starting multiprocessing\n")
results = []
tic = time.perf_counter() # Start timer
with Pool(cpu_count() * 2 - 1) as pool: # Apply the function to all the files
for result in pool.map(process, files):
if len(result) > 0:
results.append(result)
toc = time.perf_counter() # Stop timer
print(f"Finished in {toc - tic:0.4f} seconds") # Write time spent
pd.concat(results, axis=0, ignore_index=False).to_csv('features.csv', index=False) # Write csv file