-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathvoiceMusicSeparation.py
158 lines (153 loc) · 5.26 KB
/
voiceMusicSeparation.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
# -*- coding: utf-8 -*-
"""
voice music separation
--------------------------------------
reference
Po-Sen Huang / Singing-voice separation from monaural recordings using robust principal component analysis
Zhouchen Lin / The augmented lagrange multiplier method for exact recovery of corrupted low-rank matrices
---------------------------------------
used package
math/scipy/numpy
stft0.4.7 URL:https://pypi.python.org/pypi/stft/0.4.7
pymedia1.3.7.3 URL:http://www.lfd.uci.edu/~gohlke/pythonlibs/#pymedia
-------------------------------------
@author: HanSir
website: http://hansir.net
Created on Fri Apr 17 09:00:18 2015
"""
import math
import scipy
import numpy as np
# 播放音频
def playAudio(filename):
import pymedia.audio.acodec as acodec
import pymedia.muxer as muxer
import pymedia.audio.sound as sound
import time
name1 = str.split(filename, '.')
# Open demuxer first
dm = muxer.Demuxer(name1[-1].lower())
dec = None
snd = None
s = " "
f = open(filename, 'rb')
while len(s):
s = f.read(20000)
if len(s):
# 解析出最初的几帧音频数据
frames = dm.parse(s)
for fr in frames:
if dec == None:
# Open decoder
dec = acodec.Decoder(dm.streams[0])
#音频数据在 frame 数组的第二个元素中
r = dec.decode(fr[1])
if r and r.data:
if snd == None:
snd = sound.Output(r.sample_rate, r.channels, sound.AFMT_S16_LE)
snd.play(r.data)
#8.延时，直到播放完毕
while snd.isPlaying():
time.sleep(0.5)
# 收缩算子
def shrinkage(X,eps):
S = np.sign(X)*(abs(X)-1.0*eps)*(abs(X)>eps)
return S
# RPCA的IALM算法
def ialmRPCA(D, lamb=1, tol1=0.00001, tol2=0.0001, maxIter=1000):
n1,n2 = np.shape(D)
lamb = lamb/math.sqrt(max(n1,n2))
A = np.zeros((n1,n2), dtype=float)
E = np.zeros((n1,n2), dtype=float)
S = np.zeros((n1,n2), dtype=float)
# 受启发于对偶方法，Y=sgn(D)/J(sgn(D))
normDfro = np.linalg.norm(D,'fro')
normD2 = np.linalg.norm(D,2)
normDinf = np.linalg.norm(D,np.inf)
JD = max(normD2,normDinf)
Y = np.sign(D)/JD
# 根据Lin的论文
mu = 1.25/normD2
rho = 1.6
for i in range(maxIter):
U, s, Vh = scipy.linalg.svd(Y/mu+D-E,full_matrices=True)
for j in range(len(s)):
S[j][j] = s[j]
S = shrinkage(S,1/mu)
A = np.dot(np.dot(U,S),Vh)
tempE = E;
E = shrinkage(Y/mu+D-A,lamb/mu)
# update Y and mu
Y += mu*(D-A-E)
mu = rho*mu
# dispaly
stop1 = np.linalg.norm(D-A-E,'fro')/normDfro
stop2 = np.linalg.norm(E-tempE,'fro')/normDfro
print("iteration:%d, |D-A-E|_F/|D|_F:%.8f" % (i,stop1))
# stop
if(stop1<tol1 and stop2<tol2):
break;
if(i>=maxIter):
print("Max number of iter reached.")
break;
return A,E,i
# 歌声音乐分离
def voiceMusicSeparation(audio,masktype=1,lamb=1.25,gain=1.25):
import stft
# stft
specgram = stft.spectrogram(audio)
# rpca
D = abs(specgram)
angle = np.angle(specgram)
A_mag,E_mag,numiter = ialmRPCA(D,lamb)
A = A_mag*scipy.exp(angle*1j)
E = E_mag*scipy.exp(angle*1j)
# binary mask
if(masktype):
m = 1.0*(abs(E_mag)>abs(gain*A_mag))
Emask = m*specgram
Amask = specgram-Emask
else:
Emask = E
Amask = A
# istft
outputA = stft.ispectrogram(Amask)
outputE = stft.ispectrogram(Emask)
#output
wavoutA = np.array(outputA[:len(audio)],dtype=np.int16)
wavoutE = np.array(outputE[:len(audio)],dtype=np.int16)
return wavoutA, wavoutE
# 对歌曲进行歌声音乐分离（单双通道都可以处理）
def fuck(filename,masktype=1,lamb=1.25,gain=1.25):
import scipy.io.wavfile as wav
# input
fs, audio = wav.read(filename)
print("Voice Music Separation Starts...")
if(audio.shape[1]==1):
# voiceMusicSeparation
wavoutA, wavoutE = voiceMusicSeparation(audio,masktype,lamb,gain)
# output
wav.write('outputA.wav', fs, wavoutA)
wav.write('outputE.wav', fs, wavoutE)
elif(audio.shape[1]==2):
audio0 = np.array([audio[i][0] for i in range(len(audio))])
audio1 = np.array([audio[i][1] for i in range(len(audio))])
# voiceMusicSeparation
wavoutA0, wavoutE0 = voiceMusicSeparation(audio0,masktype,lamb,gain)
wavoutA1, wavoutE1 = voiceMusicSeparation(audio1,masktype,lamb,gain)
# output
wavoutA = np.array([[wavoutA0[i],wavoutA1[i]] for i in range(len(audio))])
wav.write('outputA.wav', fs, wavoutA)
wavoutE = np.array([[wavoutE0[i],wavoutE1[i]] for i in range(len(audio))])
wav.write('outputE.wav', fs, wavoutE)
else:
print("Sorry, your song is too complex to deal with.")
print("Voice Music Separation Completes.")
print("voice: outputE.wav")
print("music: outputA.wav")
##########################################################################33
#filename = 'Audio\Better Man_clip.wav'
#masktype = 1 # 1: binary mask, 0: no mask
#lamb = 1.25 # lambda in ialm_rpca
#gain = 1.25
#fuck(filename)