DSP/correlation.py at master · t4c1/DSP · 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
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
import numpy, scipy.io.wavfile,matplotlib.pyplot as plt, scipy.signal
from mysignal import play, tone

def normalize(a):
    a-=numpy.average(a)
    return a/numpy.max(numpy.abs(a))

def freq_detect(inS,inS2=None, plot_samples=1000, rate=1, full=False):
    if inS2 is None:
        inS2=inS
    inS,inS2=normalize(inS.astype(numpy.float32)),normalize(inS2.astype(numpy.float32))
    corr=numpy.correlate(inS,inS2,"full")
    if full:
        to_plot=corr
    else:
        corr=corr[corr.shape[0]/2:]
        to_plot = corr[:plot_samples]
    #corr=corr[inS2.shape[0]-1:]
    x_os = (numpy.arange(to_plot.size)-(corr.shape[0]/2*full)) / float(rate)
    plt.plot(x_os, to_plot, color="g")
    plt.show()

def radar(paket,sent,recieved):
    #paket=normalize(paket)
    sent=normalize(sent)
    recieved=normalize(recieved)

    sent_corr = numpy.correlate(sent, paket, "full")
    sent_delay=numpy.argmax(sent_corr)-sent_corr.shape[0]/2
    rec_corr = numpy.correlate(recieved, paket, "full")
    rec_delay=numpy.argmax(rec_corr)-rec_corr.shape[0]/2
    correlate = numpy.correlate(recieved, sent, "full")
    cor_delay=numpy.argmax(correlate)-correlate.shape[0]/2
    print (sent_delay,rec_delay,cor_delay)
    fig1 = plt.figure()
    #plt.plot(sent_corr, color="g")
    plt.plot(numpy.arange(sent_corr.shape[0])-sent_corr.shape[0]/2, sent_corr, color="g")
    plt.plot(numpy.arange(rec_corr.shape[0])-rec_corr.shape[0]/2, rec_corr, color="r")
    #plt.plot(numpy.arange(correlate.shape[0])-correlate.shape[0]/2, correlate, color="b")
    plt.show()

def echo_amp_find(inS,delay):
    mina=-1
    mine=float("inf")
    for a in range(0,10):
        a/=10.0
        cleaned=numpy.convolve([1]+[0]*(delay-2)+[-a], inS)
        e=numpy.sum(cleaned*cleaned)
        if e<mine:
            mina=a
            mine=e
    mina2=-1
    mine2=float("inf")
    for a in range(-5,6):
        a2=mina+a/100.0
        cleaned=numpy.convolve([1]+[0]*(delay-2)+[-a2], inS)
        e=numpy.sum(cleaned*cleaned)
        if e<mine2:
            mina2=a2
            mine2=e
    return mina2

def find_rel_amp(in1,in2, one_sided=True):
    mina=0
    if not one_sided:
        mine=float("inf")
        for a in range(0,100):
            cleaned=in1-a/10.*in2
            e=numpy.sum(cleaned*cleaned)
            if e<mine:
                mina=a
                mine=e
    mina2=-1
    mine2=float("inf")
    for a in range(-5,6):
        a2=mina+a/10.0
        cleaned=in1-a2*in2
        e=numpy.sum(cleaned*cleaned)
        if e<mine2:
            mina2=a2
            mine2=e
    mina3=-1
    mine3=float("inf")
    for a in range(-5,6):
        a3=mina2+a/100.0
        cleaned=in1-a3*in2
        e=numpy.sum(cleaned*cleaned)
        if e<mine3:
            mina3=a3
            mine3=e
    return mina3

def find_rel_amp2(in1, in2, step=1.2, steps=20, iters=2):
    prevbest=1
    for i in range(iters):
        mina=0
        mine=float("inf")
        for a in prevbest*step**((numpy.arange(0,steps)-steps/2)/float((steps/2)**i)):
            cleaned=in1-a*in2
            e=numpy.sum(cleaned*cleaned)
            print a,"\t",e
            if e<mine:
                mina=a
                mine=e
        prevbest=mina
        print "picked",mina
    return prevbest

def echo_elimination():
    sampleRate,inS = scipy.io.wavfile.read("primer_odmev.wav")
    #_,inS2 = scipy.io.wavfile.read("primer.wav")
    freq_detect(inS,inS,100000,1)#rocno delo - iskanje odmeva iz grafa korelacije
    delay=4500
    mina=echo_amp_find(inS,delay)
    mina=find_rel_amp2(inS,numpy.roll(inS,delay))
    print "amplituda odmeva:",mina
    play(numpy.hstack((inS,numpy.convolve([1]+[0]*(delay-2)+[-mina], inS))),sampleRate)#primerjava originala in brez odmevov

def beamform_1():
    # sampleRate,inS = scipy.io.wavfile.read("mic_1a.wav")
    # fns = "mic_2a.wav", "mic_3a.wav", "mic_4a.wav", "mic_5a.wav"
    # shifts=[35,70,105,140] #prebrano iz grafov, narejenih z spodnjim zakomentiranim programom. Za drugega govorca mnozi z -1
    # shifted=[]
    # cleaned=[]
    # for fn,shift in zip(fns,shifts):
    #     shift*=-1
    #     _,inS2 = scipy.io.wavfile.read(fn)
    #     mina=1.0#find_rel_amp2(inS,numpy.roll(inS2,shift)) # ni potrebno - vsi signali imajo enako amplitudo
    #     shifted.append(numpy.roll(inS2,shift)*mina)
    #     cleaned.append(numpy.roll(inS2,shift)*mina-inS)
    # #play(sum(shifted)+inS,sampleRate)
    # play(sum(cleaned),sampleRate) #najboljsi rezultat
    # #play(inS-shifted[0]+shifted[1]-shifted[2],sampleRate)

    sampleRate,inS = scipy.io.wavfile.read("mic_1a.wav")
    sampleRate,in_ = scipy.io.wavfile.read("woman1.wav")
    _,inS2 = scipy.io.wavfile.read("mic_2a.wav")
    _,inS3 = scipy.io.wavfile.read("mic_3a.wav")
    _,inS4 = scipy.io.wavfile.read("mic_4a.wav")
    _,inS5 = scipy.io.wavfile.read("mic_5a.wav")
    freq_detect(inS,in_,10000,1,True)
    freq_detect(inS2,in_,10000,1,True)
    freq_detect(inS3,in_,10000,1,True)
    freq_detect(inS4,in_,10000,1,True)
    freq_detect(inS5,in_,10000,1,True)

def beamform_2():
    sampleRate,inS = scipy.io.wavfile.read("man1.wav")
    fns = "mic_1b.wav", "mic_2b.wav", "mic_3b.wav", "mic_4b.wav", "mic_5b.wav"
    #shifts=[-21,-39,-61,-78] #prebrano iz grafov, narejenih z spodnjim zakomentiranim programom.
    shifts=[0,20,40,60,80] #prebrano iz grafov, narejenih z spodnjim zakomentiranim programom.
    shifted=[]
    cleaned=[]
    for fn,shift in zip(fns,shifts):
        shift*=-1
        _,inS2 = scipy.io.wavfile.read(fn)
        mina=1.0#find_rel_amp2(inS,numpy.roll(inS2,shift)) # ni potrebno - vsi signali imajo enako amplitudo
        shifted.append(numpy.roll(inS2,shift)*mina)
        cleaned.append(numpy.roll(inS2,shift)*mina-inS)
    play(sum(shifted),sampleRate)
    #play(sum(cleaned),sampleRate) #najboljsi rezultat
    #play(inS-shifted[0]+shifted[1]-shifted[2],sampleRate)

    # sampleRate,inS = scipy.io.wavfile.read("man1.wav")
    # _,inS1 = scipy.io.wavfile.read("mic_1b.wav")
    # _,inS2 = scipy.io.wavfile.read("mic_2b.wav")
    # _,inS3 = scipy.io.wavfile.read("mic_3b.wav")
    # _,inS4 = scipy.io.wavfile.read("mic_4b.wav")
    # _,inS5 = scipy.io.wavfile.read("mic_5b.wav")
    # freq_detect(inS,inS5,10000,1,True)

if __name__ == "__main__":
    #echo_elimination()
    #beamform_1()

    for fn in "eight_mono1.wav","eight_mono2.wav","eight_mono3.wav","eight_mono4.wav","eight_mono5.wav":
            sampleRate,inS = scipy.io.wavfile.read(fn)
            freq_detect(inS[:,0],inS[:,1],10000,1,True)

    # rad=scipy.io.loadmat("radar.mat")
    # radar(numpy.ndarray.flatten(rad["paket"]),numpy.ndarray.flatten(rad["sent1"]),numpy.ndarray.flatten(rad["rec1"]))