Exploratory Searches for Sound Effects: Verification of Similarity Based on the Acoustic Features of Sound Effects

The goal of this study is to develop an exploratory search system for sound effects (SEs). SEs appreciably influence viewers’ impressions of a movie; thus, SE editors must skillfully select the most appropriate SEs for a given scene. However, existing SE search methods have three main difficulties, involving the diversity of SE purposes, the representation of sound using text, and the conceptualization of SEs for a given scene. These difficulties lead to inefficient SE searches, because the SE editor must perform repeated searches. To solve the problems, this paper proposes a framework to define similarities among SEs with three types of features: context, acoustic features, and symbols of onomatopoeia. As the first step in this study, SEs were clustered based on their acoustic features. The relationships between the clustered SEs and their onomatopoeia characteristics are acquired through subjective evaluation experiments. As a result, it was confirmed that the classifications could help determine the content of the SE.