Subject Selection Model With Heuristic Filtering for Multi-User Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Voice assistants and smart cameras face challenges in environments with multiple users, struggling to detect and prioritize commands from the correct individual or track specific people effectively in crowded settings.
Innovation Solution
Implementing a series of rules and a machine learning model for subject selection, with memory reset mechanisms based on heuristic comparisons, to prioritize operations on selected subjects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a machine learning model is used to select subjects, then subject selection accuracy is improved, but device complexity increases
Solution Approach 1:
The system divides subject selection into multiple independent stages: first applying simple heuristic rules to filter candidates, then applying the machine learning model only to the remaining candidates. This segmentation reduces the computational burden on the ML model and overall system complexity while maintaining high selection accuracy.
Solution Approach 2:
The system performs preliminary filtering using heuristic rules before invoking the machine learning model. This preliminary action reduces the number of subjects the ML model needs to evaluate, thereby reducing computational complexity while preserving the accuracy benefits of using an ML model.
2Adaptability or versatility
If multiple subjects are detected in an environment, then system adaptability is improved, but difficulty of detecting and measuring increases
Solution Approach 1:
The system segments the detection process into distinct phases: first detecting all subjects using input devices, then applying heuristic filters to reduce the set, and finally using the ML model to make final selections. This segmentation makes the system adaptable to multiple subjects while managing detection complexity through staged processing.
Solution Approach 2:
The system introduces intermediary filtering stages between raw detection and final selection. Heuristic rules act as intermediaries that process detected subjects before they reach the ML model, reducing the complexity of measuring and distinguishing between multiple subjects in complex environments.
3Device complexity
If heuristic rules are applied for subject selection, then device complexity is reduced, but subject selection accuracy deteriorates
Solution Approach 1:
The system merges simple heuristic rules with machine learning models in a hybrid architecture. The heuristics provide simple, low-complexity filtering while the ML model adds sophisticated selection capabilities. This combination achieves high accuracy without requiring the entire system to be complex.
Solution Approach 2:
The system applies partial action by using simple heuristics for initial filtering and reserving complex ML-based selection for cases where heuristics are insufficient. This partial application of complexity only where needed achieves high accuracy while keeping overall device complexity manageable.
4Measurement precision
If a hybrid rule-based and machine learning approach is used, then subject selection accuracy is improved, but device complexity increases
Solution Approach 1:
The hybrid approach is segmented into distinct processing stages: rule-based filtering stage followed by an ML-based selection stage. This segmentation allows the system to leverage the simplicity of rules and the accuracy of ML models while managing complexity through staged execution rather than a monolithic complex system.
Solution Approach 2:
The system performs preliminary filtering using simple rules before applying the more complex machine learning model. This preliminary action reduces the input size to the ML model, thereby reducing the effective complexity required for accurate subject selection while maintaining the accuracy benefits of the hybrid approach.
Data Source
AI summary
The present disclosure generally relates to selecting objects. Some techniques are for selecting people in accordance with some embodiments. Other techniques are for resetting memory of a subject selection model in accordance with some embodiments.