X_test_wav = crop.crop_list_arrays(X_test_wav, 1000, 0.05)
print(X_train_wav)
max_length = max([len(x) for x in X_train_wav] + [len(x) for x in X_test_wav])
print("normalizing...")
X_train_wav = [wav_utils.normalize(x) for x in X_train_wav]
X_test_wav = [wav_utils.normalize(x) for x in X_test_wav]
print("padding to %d..." % max_length)
X_train_wav = wav_utils.pad_left(X_train_wav, max_length)
X_test_wav = wav_utils.pad_left(X_test_wav, max_length)
print("computing mfsc...")
X_train_mfsc = features.mfsc_matrix(X_train_wav, fft_size, window_step, num_filters, fs)
X_test_mfsc = features.mfsc_matrix(X_test_wav, fft_size, window_step, num_filters, fs)
X_train = X_train_mfsc
X_test = X_test_mfsc
# convert class vectors to binary class matrices
Y_train = np_utils.to_categorical(y_train, nb_classes)
Y_test = np_utils.to_categorical(y_test, nb_classes)
print('Build model...')
conv_features = 64
model = Sequential()
def cnn(train_speakers, test_speakers):
# the data, shuffled and split between tran and test sets
(X_train_wav, y_train), (X_test_wav, y_test) = cross_validation.load_data_set(train_speakers, test_speakers)
shuf = np.random.permutation(len(X_train_wav))
print(shuf)
X_train_wav = [X_train_wav[i] for i in shuf]
y_train = [y_train[i] for i in shuf]
'''
X_train_wav = X_train_wav[0:10]
y_train = y_train[0:10]
X_test_wav = X_test_wav[0:10]
y_test = y_test[0:10]
'''
X_train_wav = crop.crop_list_arrays(X_train_wav, 1000, 0.05)
X_test_wav = crop.crop_list_arrays(X_test_wav, 1000, 0.05)
print(X_train_wav)
max_length = max([len(x) for x in X_train_wav] + [len(x) for x in X_test_wav])
print("normalizing...")
X_train_wav = [wav_utils.normalize(x) for x in X_train_wav]
X_test_wav = [wav_utils.normalize(x) for x in X_test_wav]
print("padding to %d..." % max_length)
X_train = wav_utils.pad_middle(X_train_wav, max_length)
X_test = wav_utils.pad_middle(X_test_wav, max_length)
assert not np.any(np.isnan(X_train))
print("computing mfsc...")
X_train = features.mfsc_matrix(X_train, fft_size, window_step, num_filters, fs)
X_test = features.mfsc_matrix(X_test, fft_size, window_step, num_filters, fs)
assert not np.any(np.isnan(X_train))
print(X_train)
print('X_train shape:', X_train.shape)
print(X_train.shape[0], 'train samples')
print(X_test.shape[0], 'test samples')
# convert class vectors to binary class matrices
Y_train = np_utils.to_categorical(y_train, nb_classes)
Y_test = np_utils.to_categorical(y_test, nb_classes)
model = Sequential()
model.add(Convolution2D(nb_filter=32,
nb_row=4,
nb_col=4,
input_shape=(X_train.shape[1], X_train.shape[2], X_train.shape[3])))
model.add(Activation('relu'))
model.add(Convolution2D(nb_filter=32,
nb_row=4,
nb_col=4))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 4)))
model.add(Activation('relu'))
model.add(Convolution2D(nb_filter=32,
nb_row=4,
nb_col=4))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Activation('relu'))
model.add(Convolution2D(nb_filter=32,
nb_row=4,
nb_col=4))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Activation('relu'))
model.add(Convolution2D(nb_filter=32,
nb_row=4,
nb_col=4))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Activation('relu'))
'''
model.add(Convolution2D(nb_filter=128,
nb_row=3,
nb_col=3))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Activation('relu'))
'''
print(model.summary())
model.add(Flatten())
model.add(Dense(128))
model.add(Activation('relu'))
model.add(Dense(nb_classes))
model.add(Activation('softmax'))
sgd = SGD(lr=0.0001, decay=1e-6, momentum=0.9, nesterov=True)
model.compile(loss='categorical_crossentropy', optimizer=sgd)
model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch,
show_accuracy=True, verbose=1, validation_data=(X_test, Y_test))
score = model.evaluate(X_test, Y_test, show_accuracy=True, verbose=0)
print('Test score:', score[0])
print('Test accuracy:', score[1])
return (score[0], score[1])
