Video Search Using Semantic Pattern Matching

The patent introduces meta-data as an intermediary between the video content and the search query. Instead of directly searching video content or relying on text keywords, the system uses meta-data (including index tags and semantic patterns) extracted from video clips as a mediator. This allows the search to be performed on structured representations of video content rather than raw video data or limited text keywords, thereby improving search accuracy while maintaining system simplicity.

The patent replaces the traditional text-based keyword matching mechanism with a semantic pattern matching mechanism. Instead of relying on text search algorithms, the system extracts semantic patterns from video content (such as object relationships, actions, and scene descriptions) and compares these patterns to query patterns. This substitution enables the system to understand video content semantically rather than just matching text strings, significantly improving search effectiveness across different languages.

The patent segments the video content into discrete clips and extracts meta-data from each clip independently. The video database is organized as a collection of video clips with associated meta-data, where each clip has its own index tags and semantic patterns. This segmentation allows the search system to process and compare individual clip characteristics rather than analyzing entire videos, reducing the computational complexity while maintaining high search effectiveness.

The patent performs preliminary extraction and organization of meta-data from video content during the database building phase. Index tags and semantic patterns are pre-computed and stored with each video clip before the actual search operation. This preliminary action means that during search operations, the system only needs to compare pre-extracted meta-data rather than analyzing video content in real-time, significantly reducing the complexity of the search process while maintaining high effectiveness.

The patent extracts only the essential meta-data elements (index tags and semantic patterns) from the comprehensive MPEG-7 content description framework. Instead of implementing the full MPEG-7 standard with its complex hierarchical structure, the system extracts and utilizes only the critical components needed for video search: index tags for content identification and semantic patterns for meaning representation. This extraction approach maintains content description completeness for search purposes while avoiding the file structure complexity of the full MPEG-7 standard.

The patent applies local quality by tailoring the meta-data structure specifically for video search functionality rather than implementing a universal content description standard. The meta-data includes index tags and semantic patterns that are locally optimized for search operations, rather than providing a comprehensive description suitable for all possible multimedia operations. This localized approach reduces file structure complexity while maintaining the necessary descriptive quality for video search.