Video Segmentation and Speech Analysis for Metadata Annotation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for video material classification, such as those using speech transcripts and visual features, face challenges in effectively categorizing video segments, especially when dealing with silent segments and varying key frames, leading to inefficiencies in metadata annotation and retrieval.

Innovation Solution

A method involving video segmentation, speech-to-text analysis, key frame identification, and database matching using color histogram similarity, along with ontology-based metadata cleaning, to accurately annotate and categorize video segments, ensuring efficient metadata assignment and retrieval.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If speech transcript and visual features are used for video classification, then categorization accuracy is improved, but processing time and complexity increase

Engineering Contradiction:
Improvecategorization accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The video is divided into segments based on detected scene changes, allowing the system to process only relevant portions rather than entire videos. This segmentation reduces processing time while maintaining categorization accuracy by focusing computational resources on distinctive segments that contain classification information.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Key frames are extracted from video segments to represent the essential visual content. Instead of analyzing all frames, the system identifies and processes only the most representative key frames, significantly reducing processing time while preserving the information needed for accurate categorization.

Inventive Principle:
Principle #2Taking out (Extraction)

2Measurement precision

If comprehensive video analysis is performed on all segments, then metadata accuracy is improved, but computational resources and processing time increase

Engineering Contradiction:
Improvemetadata accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

Scene changes are detected and segments are identified before detailed analysis is performed. This preliminary action allows the system to pre-process the video structure, marking segments that require detailed analysis versus those that can be handled more simply, thereby reducing overall computational complexity while maintaining metadata accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system performs comprehensive analysis only on segments containing scene changes, while applying simplified processing to other segments. This partial action approach focuses computational resources where they are most needed (at scene transitions) while reducing overall complexity by avoiding redundant comprehensive analysis throughout the entire video.

Inventive Principle:
Principle #16Partial or excessive action

3Productivity

If key frame extraction and database matching are performed for all video segments, then retrieval efficiency is improved, but processing overhead increases

Engineering Contradiction:
Improveretrieval efficiencyVSAvoidprocessing overhead
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The video is segmented at scene changes, and key frame extraction is performed only on segments containing scene changes. This segmentation strategy reduces the total number of key frames that need to be extracted and matched against the database, improving retrieval efficiency by focusing on distinctive segments while reducing processing overhead from redundant operations on uniform segments.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP2122623B1Analysing video material
Publication Date: 2018.09.12 BRITISH TELECOM PLC
  • EP2122623B1 patent drawingFigure 1
  • EP2122623B1 patent drawingFigure 2
  • EP2122623B1 patent drawing

AI summary

Video material is divided into temporal segments. Each segment is examined to.determine whether the soundtrack of the segment contains -speech sufficient for analysis and if so, metadata are generated based on analysis of the speech. If not, the segment is analysed by comparing frames thereof with those of stored segments that already have metadata assigned to them. One then assigns to the segment under consideration stored metadata associated with one or more stored segments that are similar.