Automatic Subtitle Placement Through Active Speaker Identification in Multimedia Documents

Subtitle plays an important role in transmitting the multimedia informational content and represents the most prominent solution to substitute the audio channel in the context of deaf and hearing impaired people (HIP). In this paper we introduce a novel dynamic subtitle positioning system designed to increase the accessibility of HIP to video documents. We propose to dynamically position each textual segment in the near vicinity of the active speaker, which allows the viewer to follow the visual semantic content while simultaneously reading the subtitles. The framework exploits a rich set of technologies including: video structuring into shots and scenes, active speaker recognition, visual semantic analysis in order to position the scripts at suitable locations while increasing the HIP audience level of content understanding. The objective experimental evaluation conducted on 20 video files demonstrates the effectiveness and usefulness of the proposed approach.