Break out the Tissues: Finding Sad Scenes in Movies Using Computer Audition
For our senior conference project, we made a computer program that determines whether or not a scene in a movie is sad. To make this classification, our program uses only the audio track from the movie scene. We create a model of each movie's audio data using aspects of sound that computers can analyze (for example, pitch). We then attempt to determine whether a sound clip is sad or not sad by looking at sound clips that are most similar to it. For example, if sound clip 'x' is most similar to a sad scene, our program informs us that 'x' is a sad scene.
We're interested in this project because it could help create a mood-based search engine for movies. If trained on other emotions, our program could analyze a bunch of movies, and return the movie with the highest number of scenes containing the desired emotion.
Meggie Ladlow '09
Adam Yie '09