Awanta

ondemand/service.py

@@ -31,8 +31,9 @@ log.basicConfig(format='[%(asctime)s] %(message)s', level=log.INFO)
 AUDIOS_PATH = '/tmp'
 AHEAD_TIME_AUDIO_TOLERANCE = 2 # second
 MAX_SEGMENT_THREADS = 4
-THRESHOLD = 8
-TOLERANCE = 0.7
+THRESHOLD = 10
+SEGMENTS_TOLERANCE_RATE = 0.6
+FALL_TOLERANCE_SEGMENTS = 1
 
 # Modos de procesamiento de queue
 #  - QUEQUE_SINGLE: procesa solo un segmento a la vez
@@ -167,7 +168,9 @@ def process_segment(item, audios=None):
     try:
         audio = mutagen.mp3.MP3(filename)
         segments_needed = int(round(float(audio.info.length) / float(5)))
-        segments_needed = int(round(segments_needed * TOLERANCE))
+        segments_needed = int(round(
+            segments_needed * SEGMENTS_TOLERANCE_RATE
+        ))
     except Exception as ex:
         log.error('file {} is not an mp3'.format(audio))
         log.error(str(ex))
@@ -203,8 +206,13 @@ def process_segment(item, audios=None):
                     'offset': match['offset']
                 })
                 values.append(str(match['confidence']))
+
             except KeyError as ex:
-                log.error(str(ex))
+                the_key = str(ex)
+                if the_key == 'confidence':
+                    log.warning('Invalid confidence')
+                else:
+                    log.warning(the_key)
 
             ts += match['length'] / 1000
 
@@ -235,8 +243,9 @@ def process_segment(item, audios=None):
 def find_repetitions(results, segments_needed=2):
     found_counter = 0
     found_index = None
-    seconds_needed = 9
     expect_space = False
+    expect_recover = False
+    last_value_in_threshold_index = -1
     found = []
 
     if segments_needed < 1:
@@ -246,11 +255,34 @@ def find_repetitions(results, segments_needed=2):
         if not expect_space:
             if result['confidence'] >= THRESHOLD:
                 found_counter += 1
+                last_value_in_threshold_index = index
                 if found_index is None:
                     found_index = index
+                if expect_recover:
+                    expect_recover = False
+
+            elif FALL_TOLERANCE_SEGMENTS:
+                if not expect_recover:
+                    if last_value_in_threshold_index != -1:
+                        """ Solo cuando ya haya entrado por lo menos
+                        un valor en el rango del threshold, es cuando
+                        se podrá esperar un valor bajo """
+                        expect_recover = True
+                        found_counter += 1
+                    else:
+                        pass
+                else:
+                    """ Si después de haber pasado tolerado 1 elemento
+                    vuelve a salir otro fuera del threshold continuo,
+                    entonces ya se da por perdido """
+                    found_counter = 0
+                    found_index = None
+                    expect_recover = False
+
             else:
                 found_counter = 0
                 found_index = None
+                expect_recover = False
         else:
             if result['confidence'] <= THRESHOLD:
                 expect_space = False