AudioClassification/extract_features.py at master · UofS-CEPHIL/AudioClassification · GitHub

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
import librosa.display
import matplotlib.pyplot as plt
import glob
import os
import numpy as np
import tensorflow as tf
figure_dir = "./figures/"
data_dir = "./data/"
dir_names = list(["CoughClip", "HiccupClip", "SneezeClip", "SnoringClip", "BabyCryClip"])
audio_file_ext="*.wav"
tf_record_ext = "*.tfrecords"
vggish_dir = "./features/"


def dump_vggish_features():
    for dir_name in dir_names:
        count = 0
        label = dir_names.index(dir_name)
        for file in glob.glob(os.path.join(data_dir, dir_name, audio_file_ext)):
            command = "python ./VggishFeatures/vggish_inference_demo.py"
            param1 = " --wav_file {0}".format(file)
            store_tfrecords_path = vggish_dir + dir_name + "/" + file.split("/")[-1].split(".")[0] + ".tfrecords"
            print(store_tfrecords_path)
            param2 = " --tfrecord_file {0}".format(store_tfrecords_path)
            param3 = " --pca_params ./VggishFeatures/vggish_pca_params.npz"
            param4 = " --checkpoint ./VggishFeatures/vggish_model.ckpt"
            os.system(command+param1+param2+param3+param4)
            count += 1
            if count > 1:
                break

def load_vggish_features():
    for dir_name in dir_names:
        label = dir_names.index(dir_name)
        files = list()
        count = 0
        for file in glob.glob(os.path.join(vggish_dir, dir_name, tf_record_ext)):
            # create file queue
            files.append(file)
            record_iterator = tf.python_io.tf_record_iterator(path=file)

            for string_record in record_iterator:
                example = tf.train.SequenceExample()
                example.ParseFromString(string_record)
                # 10 seconds -> 10 steps for each step 128 d
                print(example.feature_lists.feature_list["audio_embedding"].feature[0].bytes_list.value)
            count += 1
            if count >= 1:
                break
        break


dump_vggish_features()
load_vggish_features()


def plot_spectrum_features():
    for dir_name in dir_names:
        count = 0
        label = dir_names.index(dir_name)
        for fn in glob.glob(os.path.join(data_dir, dir_name, audio_file_ext)):
            sound_file, sr = librosa.load(fn)
            plt.figure(figsize=(12, 8))
            # wave plot
            plt.subplot(4, 2, 1)
            librosa.display.waveplot(sound_file, sr=sr)
            plt.title('Wave spectrogram (Amplitude)')

            # feature extraction
            # STFT power spectrum
            plt.subplot(4, 2, 2)
            STFT = librosa.amplitude_to_db(librosa.stft(sound_file), ref=np.max)  # db fen bei
            print(STFT.shape)
            librosa.display.specshow(STFT, y_axis='log')   #or y_axis= linear; log
            plt.colorbar(format='%+2.0f dB')
            plt.title('Log-frequency power spectrogram')

            # CQT
            plt.subplot(4, 2, 3)
            CQT = librosa.amplitude_to_db(librosa.cqt(sound_file, sr=sr), ref=np.max)
            print(CQT.shape)
            librosa.display.specshow(CQT, y_axis='cqt_hz')
            plt.colorbar(format='%+2.0f dB')
            plt.title('Constant-Q power spectrogram (Hz)')

            # tempogram
            plt.subplot(4, 2, 4)
            Tgram = librosa.feature.tempogram(y=sound_file, sr=sr)
            librosa.display.specshow(Tgram, x_axis='time', y_axis='tempo')
            plt.colorbar()
            plt.title('Tempogram')
            plt.tight_layout()

            plt.subplot(4, 2, 5)
            MFCCS = librosa.feature.mfcc(y=sound_file, sr=sr, n_mfcc=20)
            print(MFCCS.T.flatten().shape)
            librosa.display.specshow(MFCCS, x_axis='time')
            plt.colorbar()
            plt.title('MFCC')
            plt.tight_layout()

            plt.subplot(4, 2, 6)
            ROLLOFF = librosa.feature.spectral_rolloff(y=sound_file, sr=sr)
            plt.semilogy(ROLLOFF.T, label='Roll-off frequency')
            plt.ylabel('Hz')
            plt.xticks([])
            plt.xlim([0, ROLLOFF.shape[-1]])
            plt.title('spectral rolloff')

            plt.subplot(4, 2, 7)
            Cent = librosa.feature.spectral_centroid(y=sound_file, sr=sr)
            plt.semilogy(Cent.T, label='Spectral centroid')
            plt.ylabel('Hz')
            plt.xticks([])
            plt.xlim([0, Cent.shape[-1]])
            plt.legend()

            plt.subplot(4, 2, 8)
            Mel = librosa.feature.melspectrogram(y=sound_file, sr=sr, n_mels=128,fmax=8000)
            librosa.display.specshow(librosa.power_to_db(Mel,ref=np.max),y_axis='mel', fmax=8000,x_axis='time')
            plt.colorbar(format='%+2.0f dB')
            plt.title('Mel spectrogram')
            plt.tight_layout()

            store_full_path = figure_dir + dir_name + "/" + fn.split("/")[-1].split(".")[0]+".png"
            print("Store position: ", store_full_path)
            plt.savefig(store_full_path)

            count += 1
            if count > 1:
                break