feat: [LoudFE] keep only N top loudness Features

add LoudAudioFeatureExtractor._num_features (set in init, default: 5)

add _keep_num -- takes list of loudnesses and keeps top N (default: 5)

pipeline/feature_extractors.py

@@ -200,12 +200,14 @@ class LoudAudioFeatureExtractor(FeatureExtractor):
 
      teardown() is used to clean up temporary files created during setup (if specified by config)
     """
-    def __init__(self, input_files=None, config=None):
+    _CONFIG_DEFAULT_NUM_FEATURES = 5 # keep the top 5 loudnesses
+    def __init__(self, input_files=None, config=None, num_features=_CONFIG_DEFAULT_NUM_FEATURES):
         if not input_files:
             raise ValueError("No input files provided!")
         self.input_files = input_files
         self.config = config
         self.features = []
+        self._num_features = num_features
 
     def _audio_file_from_path(self, path: str) -> str:
         """Return the audio file path given a video file path
@@ -249,6 +251,10 @@ class LoudAudioFeatureExtractor(FeatureExtractor):
 
         return loudness_features
 
+    def _keep_num(self, loudnesses, num=_CONFIG_DEFAULT_NUM_FEATURES) -> list:
+        """Keep the top n loudnesses (default: 5)"""
+        return sorted(loudnesses, key=lambda x: x[1], reverse=True)[:num]
+
     def setup(self):
         """extract audio from video files to be processed by pyloudnorm
 
@@ -266,7 +272,8 @@ class LoudAudioFeatureExtractor(FeatureExtractor):
         for file in self.input_files:
             audio_file = self._audio_file_from_path(file.path)
             loudnesses = self._loudnorm(audio_file)
-            for time, loudness in loudnesses:
+            top_loudnesses = self._keep_num(loudnesses, self._num_features)
+            for time, loudness in top_loudnesses:
                 self.features.append(Feature(interval=Interval(start=time, duration=0.500),
                                              source=file, feature_extractor="loudness",
                                              score=loudness))