Edge Classification Accuracy via Auto-Feature Discriminator Offset Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Edge and fog-based classification systems face accuracy issues due to reliance on global models that may not correlate well with local environments, leading to inaccurate results, especially in varying conditions such as temperature, weather, and altitude, which affect the movement of tracked objects like animals.
Innovation Solution
The implementation of an auto-feature discriminator (AFD) that generates a local-based classification model by adjusting thresholds and using moving averages based on local data, combining it with a pattern matching engine (PME) based global model to improve accuracy and adapt to local conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a global model is used for classification, then the system structure is simple, but the classification accuracy decreases due to poor correlation with local environments
Solution Approach 1:
The patent divides the classification system into two independent components: a global model (PME) that provides base classification and a local model (AFD) that provides environment-specific adjustments. This segmentation allows each model to specialize - the global model handles general patterns while the local model handles environment-specific variations - thereby improving overall accuracy without requiring a completely new complex system.
Solution Approach 2:
The patent merges the global model (PME) and local model (AFD) into a unified classification system where both models work together. The AFD generates offsets based on local environmental data that are then applied to adjust the PME classifications. This combining approach integrates the strengths of both global knowledge and local adaptation, achieving higher accuracy while maintaining a manageable system structure through modular integration.
2Use of energy by moving object
If edge processing is implemented to reduce transmission power consumption, then energy efficiency improves, but classification accuracy decreases due to limited processing capabilities
Solution Approach 1:
The patent applies partial action by implementing only the necessary processing at the edge device. Instead of performing complete complex analysis locally, the system uses simplified AFD processing that generates adjustments based on local environmental features. This partial processing approach reduces energy consumption while maintaining sufficient accuracy for the specific application needs.
Solution Approach 2:
The patent introduces an intermediary approach where the AFD acts as a mediator between the global model and the raw sensor data. The AFD processes local environmental information and generates offsets that modify the global model's classifications. This intermediary layer enables edge processing to achieve higher accuracy by bridging the gap between simple edge computing and complex global models.
3Measurement precision
If local environmental factors are considered in classification, then classification accuracy improves, but the system complexity increases
Solution Approach 1:
The patent applies local quality by creating environment-specific adjustments tailored to local conditions. The AFD analyzes local environmental features (temperature, humidity, altitude) and generates corresponding offsets that are specific to each location. This allows the system to adapt to local quality variations without redesigning the entire classification system, maintaining simplicity while improving local accuracy.
Solution Approach 2:
The patent uses parameter changes by modifying the classification parameters (offsets) based on local environmental conditions. Instead of changing the fundamental classification approach, the system adjusts the parameters of the global model according to local factors. This parameter adjustment approach improves accuracy for local conditions while keeping the overall system structure simple and manageable.
Data Source
AI summary
Methods, apparatus, systems and articles of manufacture to improve accuracy of a fog/edge-based classifier system are disclosed. An example apparatus includes a transducer to mounted on a tracked object, the transducer to generate data samples corresponding to the tracked object; a discriminator to: generate a first classification using a first model based on a first calculated feature of the first data samples from the transducer, the first model corresponding to calculated features determined from second data samples, the second data samples obtained prior to the first data samples; generate an offset based on a difference between a first model feature the first model and a second model feature of a second model, the second model being different than the first model; and adjust the first calculated feature using the offset to generate an adjusted feature; a pattern matching engine to generate a second classification using vectors corresponding to the second model based on the adjusted feature; and a counter to, when the first classification matches the second classification, increment a count.