Activity Recognition Using Contextual Graph Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current activity recognition technologies are inefficient for consumer devices due to high computation time and inability to account for contextual circumstances, such as distinguishing between a dance and a fight.
Innovation Solution
The use of contextual scoring techniques applied to known activity graphs, which generate temporal features from digital representations of observed activities and calculate similarity scores based on device context criteria.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If directed acyclic graphs are used for activity recognition, then measurement precision is improved, but productivity deteriorates due to prohibitive computation time
Solution Approach 1:
The patent segments the activity recognition process into distinct phases: feature extraction, graph construction, and graph matching. By dividing the complex DAG-based recognition into manageable segments, the system reduces computational overhead while preserving accuracy. The feature extraction phase processes video data independently, and the graph matching phase compares pre-constructed activity graphs efficiently.
Solution Approach 2:
The patent applies preliminary action by pre-construcing activity graphs from video clips before the actual recognition task. The system extracts features and builds activity graphs in advance, storing them for later comparison. This preliminary processing separates the computationally intensive graph construction from the real-time recognition decision, improving overall productivity.
2Adaptability or versatility
If traditional activity recognition is used, then device complexity is reduced, but adaptability deteriorates due to inability to account for contextual circumstances
Solution Approach 1:
The patent adds a contextual dimension to traditional activity recognition by incorporating metadata about recording conditions, device orientation, and environmental factors. This transforms the recognition system from analyzing only visual content to analyzing visual content plus contextual dimensions, enabling differentiation between activities like dance and fight based on multiple factors simultaneously.
Solution Approach 2:
The patent introduces contextual metadata as an intermediary layer between the video data and the activity recognition decision. This intermediary captures device context criteria (orientation, location, time) and integrates it with visual feature analysis, allowing the system to adapt to different circumstances without fundamentally redesigning the core recognition architecture.
3Measurement precision
If comprehensive feature analysis is used, then measurement precision is improved, but use of energy worsens due to high computational overhead on consumer devices
Solution Approach 1:
The patent extracts and separates the most discriminative features from the complete video data, focusing computational resources on key indicators of activity type. By identifying and extracting only the essential features needed for differentiation (such as motion patterns, spatial relationships), the system achieves accurate recognition with reduced energy consumption compared to analyzing all video features comprehensively.
Data Source
AI summary
An activity recognition system is disclosed. A plurality of temporal features is generated from a digital representation of an observed activity using a feature detection algorithm. An observed activity graph comprising one or more clusters of temporal features generated from the digital representation is established, wherein each one of the one or more clusters of temporal features defines a node of the observed activity graph. At least one contextually relevant scoring technique is selected from similarity scoring techniques for known activity graphs, the at least one contextually relevant scoring technique being associated with activity ingestion metadata that satisfies device context criteria defined based on device contextual attributes of the digital representation, and a similarity activity score is calculated for the observed activity graph as a function of the at least one contextually relevant scoring technique, the similarity activity score being relative to at least one known activity graph.