Scene-Aware Audio-Video Graphs for Object Interaction Analysis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing technologies fail to provide a comprehensive and efficient representation of audio-video scenes that capture both spatial and temporal interactions of objects, necessitating separate processing of audio and video modalities without aligning them effectively.

Innovation Solution

A graph-based representation is generated, where nodes represent objects and edges indicate interactions, incorporating audio and video features to create a scene-aware audio-video representation, utilizing a neural network for audio-source separation and feature extraction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If separate processing of audio and video modalities is used, then processing simplicity is maintained, but representation comprehensiveness deteriorates

Engineering Contradiction:
Improveprocessing simplicityVSAvoidrepresentation comprehensiveness
Core Design Contradiction:
Device complexityVSLoss of information

Solution Approach 1:

The patent merges audio and video processing into a unified graph representation framework. Nodes represent objects from both modalities, and edges represent relationships between them, creating an integrated scene-aware representation that captures interactions between audio and video elements without requiring separate processing pipelines.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The graph representation serves multiple functions simultaneously: it represents spatial relationships between objects, temporal evolution of scenes, audio-visual alignments, and object interactions. This universal representation eliminates the need for separate processing systems while maintaining comprehensive information capture.

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Reliability

If pre and post-processing techniques are applied to align audio and video, then modality alignment is improved, but processing time increases

Engineering Contradiction:
Improvemodality alignmentVSAvoidprocessing time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The system performs alignment actions during the main processing stage rather than as separate pre-processing or post-processing steps. The graph representation is constructed to inherently capture aligned audio-visual relationships, eliminating the need for time-consuming separate alignment operations while maintaining reliable modality synchronization.

Inventive Principle:
Principle #10Preliminary action

3Loss of information

If graph-based representation with audio-video integration is used, then representation comprehensiveness is improved, but computational complexity increases

Engineering Contradiction:
Improverepresentation comprehensivenessVSAvoidcomputational complexity
Core Design Contradiction:
Loss of informationVSDevice complexity

Solution Approach 1:

The complex audio-video scene is segmented into discrete graph nodes representing individual objects and edges representing relationships between them. This segmentation allows the system to manage complexity by processing relationships between pairs of objects rather than the entire scene at once, while maintaining comprehensive representation through the complete graph structure.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4374368B1A method and system for scene-aware audio-video representation
Publication Date: 2025.12.31 MITSUBISHI ELECTRIC CORP
  • EP4374368B1 patent drawingFigure 1A
  • EP4374368B1 patent drawingFigure 1B
  • EP4374368B1 patent drawingFigure 1C

AI summary

Embodiments disclose a method and system for a scene-aware audio-video representation of a scene. The scene-aware audio video representation corresponds to a graph of nodes connected by edges. A node in the graph is indicative of the video features of an object in the scene. An edge in the graph connecting two nodes indicates an interaction of the corresponding two objects in the scene. In the graph, at least one or more edges are associated with audio features of a sound generated by the interaction of the corresponding two objects. The graph of the audio-video representation of the scene may be used to perform a variety of different tasks. Examples of the tasks include one or a combination of an action recognition, an anomaly detection, a sound localization and enhancement, a noisy-background sound removal, and a system control.