fix LPC NaN Bug

README.md

@@ -64,7 +64,9 @@ Our GUI not only has basic functionality for recording, enrollment, training and
 
 ![graph](https://github.com/ppwwyyxx/speaker-recognition/raw/master/doc/Final-Report-Complete/img/gui-graph.png)
 
-See our [demo video](https://github.com/ppwwyyxx/speaker-recognition/raw/master/demo.avi) (in Chinese) for more details.
+You should understand that real-time speaker recognition is extremely hard, because we only use corpus of about 1 second length to identify the speaker.
+Therefore the real-time system doesn't work very perfect.
+You can See our [demo video](https://github.com/ppwwyyxx/speaker-recognition/raw/master/demo.avi) (in Chinese).
 
 ## Command Line Tools
 ```sh

src/feature/LPC.py

@@ -1,7 +1,7 @@
 #!/usr/bin/env python2
 # -*- coding: UTF-8 -*-
 # File: LPC.py
-# Date: Sat Feb 21 18:41:51 2015 +0800
+# Date: Thu Mar 19 19:37:11 2015 +0800
 # Author: Yuxin Wu <[email protected]>
 
 import time
@@ -52,7 +52,9 @@ def extract(self, signal):
             frame[1:] -= frame[:-1] * self.PRE_EMPH
             feature.append(self.lpcc(frame))
 
-        return array(feature)
+        feature = array(feature)
+        feature[isnan(feature)] = 0
+        return feature
 
 @cached_func
 def get_lpc_extractor(fs, win_length_ms=32, win_shift_ms=16,