Fix window right boundary time larger than audio file max time

autovot/bin/auto_vot_extract_features.py

@@ -147,6 +147,9 @@ def textgrid2front_end(textgrid_list, wav_list, input_filename, features_filenam
             logging.error('(sox can be downloaded from http://sox.sourceforge.net)')
             problematic_files.append(wav_filename)
             continue
+        num_samples = wav_file.getnframes()
+        wav_duration = num_samples/float(wav_file.getframerate())
+        print "wav_duration=", wav_duration
         textgrid = TextGrid()
 
         # read TextGrid
@@ -161,15 +164,16 @@ def textgrid2front_end(textgrid_list, wav_list, input_filename, features_filenam
             logging.error("Either --window_tier or --vot_tier should be given.")
             exit(-1)
 
-            # check if the VOT tier is one of the tiers in the TextGrid
+        # check if the VOT tier is one of the tiers in the TextGrid
         if definitions.vot_tier in tier_names:
             tier_index = tier_names.index(definitions.vot_tier)
             # run over all intervals in the tier
             for interval in textgrid[tier_index]:
                 if (definitions.vot_mark == "*" and re.search(r'\S', interval.mark())) \
                         or (interval.mark() == definitions.vot_mark):
                     window_min = max(interval.xmin() + definitions.window_min, 0)
-                    window_max = min(interval.xmax() + definitions.window_max, textgrid.xmax())
+                    window_max = min(min(interval.xmax() + definitions.window_max, textgrid.xmax()), wav_duration)
+                    print window_max
                     new_instance = Instance()
                     new_instance.set(wav_filename, window_min, window_max, interval.xmin(), interval.xmax())
                     instances.append(new_instance)

autovot/bin/helpers/textgrid.py

@@ -137,7 +137,7 @@ def read(self, file):
         text.readline()
         for i in range(m): # loop over grids
             text.readline()
-            if text.readline().rstrip().split()[2] == '"IntervalTier"': 
+            if text.readline().rstrip().split()[2] == '"IntervalTier"':
                 # inam = text.readline().rstrip().split()[2][1:-1]
                 inam = text.readline().split('=')[1].strip().strip('"') # Joseph Keshet: handle space in the tier name
                 imin = float(text.readline().rstrip().split()[2])

autovot/code/vot_predictor/VotFrontEnd2.cpp

@@ -131,7 +131,9 @@ int main(int argc, char **argv)
 			instances.word_end[i] = (samples.size()-1)/double(sampling_rate);
 		}
 
-		if (instances.word_end[i] >= samples.size()/double(sampling_rate)) {
+		// round the wav file duration up to 3 figures after the point
+		double wav_duration = round(1000*samples.size()/double(sampling_rate))/1000.0;
+		if (instances.word_end[i] > wav_duration) {
 			LOG(ERROR) << "Word end (" << instances.word_end[i] << ") is greater than "
 			<< "the length of " << instances.file_list[i];
 			return EXIT_FAILURE;