Wireless Sensing Classifier Refinement for Rich-Scattering IoT Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wireless sensing technologies face challenges in efficiently utilizing multiple IoT devices for effective wireless sensing due to the complexity of wireless signal interactions with human activities, particularly in rich-scattering environments, necessitating an improved method for data processing and classifier refinement.
Innovation Solution
The method involves classifier probing and refinement, which includes obtaining raw measurement data, constructing input data, performing classification, perturbing input data, and re-training the classifier based on selected perturbed input data to enhance the accuracy of wireless sensing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If multiple IoT devices are deployed for wireless sensing in rich-scattering environments, then sensing coverage and data collection capability are improved, but signal interaction complexity and processing difficulty increase
Solution Approach 1:
The patent segments the complex sensing data processing task into multiple manageable steps: raw data collection from multiple devices, feature extraction, classifier probing with perturbations, and iterative refinement. This segmentation allows the system to handle complexity systematically by breaking down the overall processing pipeline into distinct stages that can be managed independently.
Solution Approach 2:
The patent implements feedback mechanisms through classifier probing and refinement. The system probes the classifier with perturbed inputs, compares outputs against reference outcomes, and uses the deviations to refine the classifier iteratively. This feedback loop enables the system to adapt to complex signal interactions by continuously learning from probing results and adjusting the classifier accordingly.
2Measurement precision
If classifier accuracy is improved through iterative refinement, then sensing precision is enhanced, but processing time and computational resources increase
Solution Approach 1:
The patent applies partial action by performing classifier probing only on selected perturbed inputs rather than exhaustively testing all possible inputs. The system identifies reference inputs with known outcomes and generates perturbations only for these critical cases, then stops refinement when sufficient accuracy is achieved. This partial approach balances precision improvement with time efficiency by avoiding unnecessary computational exhaustiveness.
Solution Approach 2:
The patent performs preliminary actions by pre-identifying reference inputs with known reference outcomes before the main classification task. This preliminary preparation allows the iterative refinement process to focus computational resources only on the most critical classification decisions, reducing overall processing time while maintaining high precision through targeted probing and refinement of key input-output mappings.
3Measurement precision
If data processing steps are increased for better classification, then detection accuracy is improved, but system complexity and implementation difficulty increase
Solution Approach 1:
The patent makes the processing framework universal by designing a multi-functional classifier probing and refinement mechanism that can handle various sensing scenarios and data types through a unified approach. The same probing and refinement process works across different IoT device configurations and environmental conditions, reducing implementation complexity by providing a single adaptable framework rather than multiple specialized processing pipelines.
Data Source
AI summary
Wireless sensing using classifier probing and refinement is described. In one example, a described method comprises: computing output analytics by a classifier based on input data constructed based on raw measurement data; mapping each output analytics to a respective mapped outcome; identifying at least one reference input data each associated with a reference output analytics and a reference mapped outcome; each reference input data being one of the input data for the classifier for which a respective reference outcome is available and is different from the reference mapped outcome; and for each reference input data for the classifier: computing a respective plurality of perturbed output analytics by the classifier, generating selected perturbed input data based on the plurality of perturbed output analytics; and re-training the classifier based on the selected perturbed input data and the associated reference outcome.