diff --git a/Trial.py b/Trial.py
@@ -9,17 +9,25 @@
 NoFart=extractFeature('./data/negative_sample_party_mono.wav')
 NoFart2=extractFeature('./data/negative_sample_voice_mono.wav')
 
-X=np.concatenate([Fart,NoFart,NoFart2],axis=1)
+Fart.shape
+NoFart.shape
+
+#X=np.concatenate([Fart,NoFart,NoFart2],axis=1)
+#Y=np.asarray([1  for i in range(Fart.shape[1])]+\
+ #       [0  for i in range(NoFart.shape[1]+NoFart2.shape[1])]).astype('float32')
+
+
+X=np.concatenate([Fart,NoFart],axis=1)
 Y=np.asarray([1  for i in range(Fart.shape[1])]+\
-        [0  for i in range(NoFart.shape[1]+NoFart2.shape[1])]).astype('float32')
+        [0  for i in range(NoFart.shape[1])]).astype('float32')
 
 VecNP=np.vstack([X,Y]).T
 VecNP.shape
 np.random.shuffle(VecNP)
 X=VecNP[:,range(40)]
 Y=VecNP[:,40]
 
-TrainSize=150000
+TrainSize=60000
 ValidationSize=X.shape[0]-TrainSize
 
 
@@ -125,7 +133,28 @@
 
 l2_model.evaluate(x=X_validation, y = Y_validation, batch_size=None, verbose=1)
 
-Predict=list(chain.from_iterable(l2_model.predict(X_Test2).tolist()))
+Predict=list(chain.from_iterable(l2_model.predict(X_validation,).tolist()))
+type(Predict)
+Real=Y_validation.tolist()
+np.corrcoef(Predict,Real)
+
+plt.clf()   # clear figure
+plt.plot(Predict,Real,'bo')
+
+plt.show()
+
+
+
+#new test
+from keras.models import load_model
+
+from itertools import chain
+
+#model2=load_model('Fart_model.h5')
+
+l2_model.evaluate(x=X_Test2, y = Y_Test2, batch_size=None, verbose=1)
+
+Predict=list(chain.from_iterable(l2_model.predict(X_Test2,).tolist()))
 type(Predict)
 Real=Y_Test2.tolist()
 np.corrcoef(Predict,Real)