def pgram(x, fs=1.0):
"""
A function to plot the periodogram of a signal x
(C) 2013 Noah D. Brenowitz
"""
from numpy import fft
ps = abs(fft.fft(x**2))**2/x.shape[0]
print(ps.max())
fs = fft.fftfreq(x.shape[0],fs)
return plt.loglog(fs, ps)
def mfft(v):
"utility that returns the modulus of the fft of v"
import scipy.fftpack as fft
s0 = fft.fft(v) ###
s0 = np.real(np.sqrt(s0 * s0.conj())) # ref spectrum
return s0
def svmexperiment():
o = open('C:\Users\Sonmaz\Desktop\gestures\output.txt','a')#call the following function with the sign file and the example between 0 and 10
for letter in letters:
xins, yins = signSection('C:\Users\Sonmaz\Desktop\gestures\graffiti-'+letter+'.txt', 0.3)
for i in [0,1,2]:
rand= random.uniform(0,len(xins))
rand=int(floor(rand))
testD.append([letter,xins.pop(rand),yins.pop(rand)])
for i in range(len(xins)):
trainD.append([letter,xins[i],yins[i]])
#===================================================================end of reading the file==========================================
for td in testD:
tmp = Gesture()
tmp.name = td[0]
l = len(td[1])
ls = (l / (n+1))
for i in range(0, n):
tmp.frame1.append([td[1][i * ls : (i+1) * ls],td[2][i * ls : (i+1) * ls]])
for fr1 in range(len(tmp.frame1)-1):
if len(tmp.frame1[fr1][0]+tmp.frame1[fr1+1][0]) < ls:
break
if len(tmp.frame1[fr1][1]+tmp.frame1[fr1+1][1]) < ls:
break
tmp.frame.append([tmp.frame1[fr1][0]+tmp.frame1[fr1+1][0],tmp.frame1[fr1][1]+tmp.frame1[fr1+1][1]])
for fr in tmp.frame:
trans=[]
trans.append(fft.fft(fr[0]))
trans.append(fft.fft(fr[1]))
sum1= 0
sum2 = 0
tmp.features.append(abs(trans[0][0]))
tmp.features.append(abs(trans[1][0]))
for i in range(len(trans[0])):
sum1 += trans[0][i]*trans[0][i]
for i in range(len(trans[1])):
sum2 += trans [1][i]*trans[1][i]
tmp.features.append(int(abs(sum1)/(2*ls - 1)))
tmp.features.append(int(abs(sum2)/(2*ls - 1)))
while len(tmp.features) < (n+1):
tmp.features.append(0)
print 'faaaaaa'
print len(tmp.features)
gesturesD.append(tmp)
for tr in trainD:
tmp = Gesture()
tmp.name = tr[0]
l = len(tr[1])
ls = l / (n+1)
for i in range(0, n):
tmp.frame1.append([tr[1][i * ls : (i+1) * ls],tr[2][i * ls : (i+1) * ls]])
for fr1 in range(len(tmp.frame1)-1):
if len(tmp.frame1[fr1][0]+tmp.frame1[fr1+1][0]) < ls:
break
if len(tmp.frame1[fr1][1]+tmp.frame1[fr1+1][1]) < ls:
break
tmp.frame.append([tmp.frame1[fr1][0]+tmp.frame1[fr1+1][0],tmp.frame1[fr1][1]+tmp.frame1[fr1+1][1]])
for fr in tmp.frame:
trans=[]
#print fr[0]
trans.append(fft.fft(fr[0]))
trans.append(fft.fft(fr[1]))
sum1= 0
sum2 = 0
tmp.features.append(abs(trans[0][0]))
tmp.features.append(abs(trans[1][0]))
for i in range(len(trans[0])):
sum1 += trans[0][i]*trans[0][i]
for i in range(len(trans[1])):
sum2 += trans[1][i]*trans[1][i]
tmp.features.append(int(abs(sum1)/(2*ls - 1)))
tmp.features.append(int(abs(sum2)/(2*ls - 1)))
print 'taaaa'
while len(tmp.features) < (n+1):
tmp.features.append(0)
print len(tmp.features)
gesturesT.append(tmp)
#============================================================================end of feature extraction
truelabels=[]
attributes=[]
labels=[]
testAtr=[]
for ge in gesturesT:
tempAtr=[]
for fe in ge.features:
tempAtr.append(fe)
attributes.append(tempAtr)
if(ge.name == 'e'):
labels.append(-1)
else:
labels.append(1)
for ge in gesturesD:
tempAtr=[]
for fe in ge.features:
tempAtr.append(fe)
testAtr.append(tempAtr)
if(ge.name == 'e'):
truelabels.append(-1)
else:
truelabels.append(1)
#print attributes
attributes = array(attributes)
print attributes
mysvm = Svm()
#print testAtr
testAtr=array(testAtr)
print 'sep'
labels=array(labels)
#labels = array([-1,1,-1,1,-1,1,-1,1,-1,1,-1,1,-1,1,-1,1,-1,1,-1,1])
mysvm.compute(attributes,labels)
print testAtr
predictedlabels = mysvm.predict(testAtr)
o.write('\n'+str(truelabels))
o.write(str(predictedlabels))
correct = 0
for la in range(len(truelabels)):
if truelabels[la] == predictedlabels[la]:
correct += 1
print correct