Upload FeatureExtraction.py
Browse files- FeatureExtraction.py +59 -0
FeatureExtraction.py
ADDED
@@ -0,0 +1,59 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/env python
|
2 |
+
# coding: utf-8
|
3 |
+
|
4 |
+
# In[ ]:
|
5 |
+
|
6 |
+
|
7 |
+
import librosa
|
8 |
+
import librosa.display
|
9 |
+
import numpy as np
|
10 |
+
import pandas as pd
|
11 |
+
import scipy
|
12 |
+
from scipy.stats import skew
|
13 |
+
|
14 |
+
import matplotlib.pyplot as plt
|
15 |
+
|
16 |
+
SAMPLE_RATE = 44100
|
17 |
+
def get_mfcc(name, path):
|
18 |
+
data, _ = librosa.core.load(path + name, sr = SAMPLE_RATE)
|
19 |
+
assert _ == SAMPLE_RATE
|
20 |
+
try:
|
21 |
+
ft1 = librosa.feature.mfcc(data, sr = SAMPLE_RATE, n_mfcc=30)
|
22 |
+
ft2 = librosa.feature.zero_crossing_rate(data)[0]
|
23 |
+
ft3 = librosa.feature.spectral_rolloff(data)[0]
|
24 |
+
ft4 = librosa.feature.spectral_centroid(data)[0]
|
25 |
+
ft5 = librosa.feature.spectral_contrast(data)[0]
|
26 |
+
ft6 = librosa.feature.spectral_bandwidth(data)[0]
|
27 |
+
ft1_trunc = np.hstack((np.mean(ft1, axis=1), np.std(ft1, axis=1), skew(ft1, axis = 1), np.max(ft1, axis = 1),
|
28 |
+
np.median(ft1, axis = 1), np.min(ft1, axis = 1)))
|
29 |
+
ft2_trunc = np.hstack((np.mean(ft2), np.std(ft2), skew(ft2), np.max(ft2), np.median(ft2), np.min(ft2)))
|
30 |
+
ft3_trunc = np.hstack((np.mean(ft3), np.std(ft3), skew(ft3), np.max(ft3), np.median(ft3), np.min(ft3)))
|
31 |
+
ft4_trunc = np.hstack((np.mean(ft4), np.std(ft4), skew(ft4), np.max(ft4), np.median(ft4), np.min(ft4)))
|
32 |
+
ft5_trunc = np.hstack((np.mean(ft5), np.std(ft5), skew(ft5), np.max(ft5), np.median(ft5), np.min(ft5)))
|
33 |
+
ft6_trunc = np.hstack((np.mean(ft6), np.std(ft6), skew(ft6), np.max(ft6), np.median(ft6), np.max(ft6)))
|
34 |
+
return pd.Series(np.hstack((ft1_trunc, ft2_trunc, ft3_trunc, ft4_trunc, ft5_trunc, ft6_trunc)))
|
35 |
+
except:
|
36 |
+
print('bad file')
|
37 |
+
return pd.Series([0]*210)
|
38 |
+
|
39 |
+
def MFCC_spectrogram(y, sr):
|
40 |
+
# 提取頻域音頻資料 MFCC spectrogram feature
|
41 |
+
melspec = librosa.feature.melspectrogram(y, sr, n_fft=1024, hop_length=512, n_mels=128)
|
42 |
+
# 轉成 log scale 已呈現頻譜資料
|
43 |
+
logmelspec = librosa.power_to_db(melspec)
|
44 |
+
# 繪圖
|
45 |
+
plt.figure()
|
46 |
+
librosa.display.specshow(logmelspec, sr=sr, x_axis='time', y_axis='mel')
|
47 |
+
plt.title('Features of unknow frog by MFCC (Mel Frquency Cepstral Coefficients)')
|
48 |
+
plt.show()
|
49 |
+
|
50 |
+
|
51 |
+
|
52 |
+
def FeatureExtraction(fname):
|
53 |
+
# 創建文件檔
|
54 |
+
test= pd.DataFrame({'fname':[fname]})
|
55 |
+
# 特徵提取
|
56 |
+
test_data = test['fname'].progress_apply(get_mfcc, path='')
|
57 |
+
test_data = test_data.values
|
58 |
+
return test_data
|
59 |
+
|