def main():
kannada = "C:\\Users\\bhuvan\\Desktop\\training set\\Kannada\\wav"
hindi = "C:\\Users\\bhuvan\\Desktop\\training set\\Hindi\\wav"
kan_files = test_gmm_win.list_files(['--dirpath=' + kannada])
hindi_files = test_gmm_win.list_files(['--dirpath=' + hindi])
nobj = gmm.GaussianMixtureModel.loadobject("C:\\Users\\bhuvan\\sonus\\gmm-object")
kan_detected = 0
hin_detected = 0
for i in range(600, 800):
a = sonusreader.SonusReader.from_file(os.path.join(kannada, kan_files[i]))
b = sonusreader.SonusReader.from_file(os.path.join(hindi, hindi_files[i]))
adata = mfcc.mfcc(a.data, samplerate=a.samplerate)
bdata = mfcc.mfcc(b.data, samplerate=b.samplerate)
if nobj.fit(adata) == 0:
kan_detected = kan_detected + 1
if nobj.fit(bdata) == 1:
hin_detected = hin_detected + 1
print '-' * 30
print 'Language', '\t\ttested files', '\t\tdetected files'
print 'Kannada' , '\t\t' + str(200), '\t\t' + str(kan_detected)
print 'Hindi', '\t\t' + str(200), '\t\t' + str(hin_detected)
print '-' * 30
def main():
kannada = "C:\\Users\\bhuvan\\Desktop\\training set\\Kannada\\wav"
hindi = "C:\\Users\\bhuvan\\Desktop\\training set\\Hindi\\wav"
kan_files = list_files(['--dirpath=' + kannada])
hindi_files = list_files(['--dirpath=' + hindi])
numfiles = min(len(kan_files), len(hindi_files))
len_train_set = int(round(0.6 * numfiles))
training_set = []
for i in range(len_train_set):
training_set.append((os.path.join(kannada, kan_files[i]),
os.path.join(hindi, hindi_files[i])))
data = []
audio1 = sonusreader.SonusReader.from_file(training_set[0][0])
audio2 = sonusreader.SonusReader.from_file(training_set[0][1])
mfcc1 = mfcc.mfcc(audio1.data, samplerate=audio1.samplerate)
mfcc2 = mfcc.mfcc(audio2.data, samplerate=audio2.samplerate)
data = np.vstack((mfcc1, mfcc2))
for example in training_set[1:]:
audio1 = sonusreader.SonusReader.from_file(example[0])
audio2 = sonusreader.SonusReader.from_file(example[1])
mfcc1 = mfcc.mfcc(audio1.data, samplerate=audio1.samplerate)
mfcc2 = mfcc.mfcc(audio2.data, samplerate=audio2.samplerate)
d = np.vstack((mfcc1, mfcc2))
data = np.vstack((data, d))
print 'done'
GMM = gmm.GaussianMixtureModel(data, 2, options={
'method':'uniform'
})
GMM.expectationMaximization()
gmm.GaussianMixtureModel.saveobject(GMM, filepath="C:\\Users\\bhuvan\\sonus\\uniform")