Repository navigation
Expand file tree
/
Copy pathaudio_manager.py
More file actions
94 lines (70 loc) · 2.65 KB
/
Copy pathaudio_manager.py
File metadata and controls
94 lines (70 loc) · 2.65 KB
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
import wave
import librosa
import numpy as np
import pandas as pd
import soundfile as sf
import system_manager
class AudioManager:
def __init__(self, path):
self.path = path
def get_bpm(self):
duration = 30
x_sr = 200
bpm_min, bpm_max = 60, 240
# 楽曲の信号を読み込む
y, sr = librosa.load(self.path, offset=38, duration=duration, mono=True)
# ビート検出用信号の生成
# リサンプリング & パワー信号の抽出
x = np.abs(librosa.resample(y, sr, x_sr)) ** 2
x_len = len(x)
# 各BPMに対応する複素正弦波行列を生成
M = np.zeros((bpm_max, x_len), dtype=complex)
for bpm in range(bpm_min, bpm_max):
thete = 2 * np.pi * (bpm / 60) * (np.arange(0, x_len) / x_sr)
M[bpm] = np.exp(-1j * thete)
# 各BPMとのマッチング度合い計算
# (複素正弦波行列とビート検出用信号との内積)
x_bpm = np.abs(np.dot(M, x))
# BPM を算出
bpm = np.argmax(x_bpm)
return bpm
def length(self):
wf = wave.open(self.path, mode='rb')
return wf.getnframes() / wf.getframerate()
@staticmethod
def labels_list_to_df(labels_list):
labels_01_list = []
for i in range(len(labels_list)):
labels_01_list.append(np.bincount([labels_list[i]]))
return pd.DataFrame(labels_01_list)
@staticmethod
def df_to_df_list(df):
df_list = []
for i in range(len(df.columns)):
df_list.append(df[i])
return df_list
def get_ignore_time(self):
data, samplerate = sf.read(self.path)
config = system_manager.get_config()['ignore_sound_range']
thres = config['thres']
amp = np.abs(data)
b = amp > thres
min_silence_duration = config['min_silence_duration']
silences = []
prev = 0
entered = 0
for i, v in enumerate(b):
t = type(prev)
if (t is int and prev == 1) or (t is np.ndarray and prev.any()) \
and not v.all(): # enter silence
entered = i
if (t is int and prev == 0) or (t is np.ndarray and not prev.all()) \
and v.any(): # exit silence
duration = (i - entered) / samplerate
if duration > min_silence_duration:
silences.append([entered / samplerate, i / samplerate])
entered = 0
prev = v
if 0 < entered < len(b):
silences.append([entered / samplerate, len(b) / samplerate])
return silences