Browse Prior Art Database

AUTOMATIC MULTILINGUAL SUBTITLE ALIGNMENT FOR VIDEO CONTENT

IP.com Disclosure Number: IPCOM000246199D
Publication Date: 2016-May-16
Document File: 8 page(s) / 827K

Publishing Venue

The IP.com Prior Art Database

Related People

Harel Cain: AUTHOR [+4]

Abstract

Presented herein is an automatic method for synchronizing subtitles to correctly fit a video program, taking into account time offsets, scaling factors and skips. This can be based on synchronizing one subtitle file to fit another which is known to match the video, or even to match the subtitle file directly against the video. Using such an automatic system, broadcasters can leverage freely available subtitles and offer them as an additional service to their subscribers, after automatically ensuring that they fit the version of video they are delivering.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 23% of the total text.

Page 01 of 8

AUTOMATIC MULTILINGUAL SUBTITLE ALIGNMENT FOR VIDEO CONTENT

AUTHORS:

 Harel Cain Michal Devir Yaron Sella Eran Iceland

CISCO SYSTEMS, INC.

ABSTRACT

    Presented herein is an automatic method for synchronizing subtitles to correctly fit a video program, taking into account time offsets, scaling factors and skips. This can be based on synchronizing one subtitle file to fit another which is known to match the video, or even to match the subtitle file directly against the video. Using such an automatic system, broadcasters can leverage freely available subtitles and offer them as an additional service to their subscribers, after automatically ensuring that they fit the version of video they are delivering.

DETAILED DESCRIPTION

     Television broadcasters and content providers operating in an international market or in countries with a high percentage of foreign language speakers could benefit from offering their subscribers or customers subtitling for video content in many languages, as many as a few dozen.

    Free / open subtitle websites contain such subtitles in many languages. These subtitles are essentially released into the public domain or with a license that could let anyone use them under proper attribution. However, because they were created using videos from dubious sources (such as file sharing networks and others), they are synchronized/aligned with many different versions of the videos.

    Once there is an automatic way to align two different subtitles (of different languages even) against each other, the broadcaster can use a formal, verified-to-match- the-video subtitle it has produced professionally and align other subtitles, available online,

Copyright 2016 Cisco Systems, Inc.

1


Page 02 of 8

against it. Alternatively, these subtitles can be aligned against the audio component of the

broadcaster's formal version of the video.

    The various subtitles, even if correct in terms of contents, can be misaligned against the broadcaster's legal, formal version of the content (the one about to be broadcast) for at least three different reasons:

    1. Offset - the subtitles begin with a certain time offset, either positive or negative, relative to the version of the video.

    2. Scaling factor - the subtitles use a different timebase / frame rate than the one used by the version of the video, such that every 1 second in the video corresponds to some number λ = 1 of seconds in the subtitles. Typically, λ is some number close to 1, but slightly higher or lower than 1. Such a scaling factor can lead to a gap of a minute or more over the course of a feature film for example, between the audio and the subtitles.

    3. Skips - these occur whenever, in either the subtitles or the video, a certain time period is "skipped" (or, equivalently, "inserted" into the other file). Skips occur many times among various copies of the video, and involve short periods of video being skipped in relation to each other. The reasons for skips are manual editing (by amateurs...