AI Sign Language Video Generation From Subtitles for Accessible Media
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Hearing-impaired individuals face challenges in comprehending non-verbal sounds in audio-video content, as subtitles often fail to convey these sounds, and existing solutions like picture-in-picture gesture language videos consume screen space and require specialized expertise, adding costs.
Innovation Solution
A data storage device generates sign language videos from subtitles using artificial intelligence, allowing these videos to be superimposed or presented as a separate track, enhancing media experience without distracting from the main content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If picture-in-picture gesture language video is used, then accessibility for hearing-impaired viewers is improved, but screen space is consumed and device complexity increases
Solution Approach 1:
The patent uses AI to automatically generate gesture language videos from subtitles, creating a copy of the textual information in visual form. This eliminates the need for manual creation of gesture videos, reducing device complexity while maintaining accessibility benefits
Solution Approach 2:
The system changes the format parameter of subtitle information from text to video by using AI generation. This parameter transformation allows the same information to be delivered in a visually accessible format without requiring complex manual intervention
2Adaptability or versatility
If picture-in-picture gesture language video is used, then accessibility for hearing-impaired viewers is improved, but storage requirements increase
Solution Approach 1:
Instead of storing separate gesture video files, the system generates gesture videos on-demand from stored subtitle data. This copying approach reduces storage requirements while maintaining the ability to provide accessible content
Solution Approach 2:
The system performs preliminary AI model training and stores the trained model, then uses it to generate gesture videos when needed. This preliminary action reduces the need for storing multiple pre-generated gesture video variants
3Manufacturing precision
If manual gesture language video creation is used, then accuracy of gesture representation is improved, but manufacturing cost and time increase
Solution Approach 1:
The system copies textual subtitle information and transforms it into gesture video format using AI, eliminating the need for manual gesture creation while maintaining reasonable accuracy through the AI model's learning capabilities
Solution Approach 2:
The patent replaces the mechanical process of manual gesture video creation with an automated AI-based system. This substitution dramatically reduces production time and cost while maintaining acceptable accuracy through the AI model's ability to learn from training data
Data Source
AI summary
A data storage device and method are provided for gesture generation and management. In one embodiment, a data storage device is provided comprising a memory and one or more processors. The one or more processors, individually or in combination, are configured to: extract subtitles from a video stored in the memory; generate a gesture video from the subtitles; create a combined video comprising the generated gesture video combined with the video; and store the combined video in the memory. Other embodiments are provided.