Home Apparatus State Detection via Feature Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting apparatus states in homes require extensive learning and are prone to errors, especially when unknown apparatuses are introduced, and anomaly detection systems struggle with accurately identifying unusual activities without causing false alarms.
Innovation Solution
An apparatus state detector that measures physical quantities, calculates feature quantities, and references a stored dictionary to identify apparatus states, combined with a living persons' anomaly detector that uses wireless communication to determine activity levels and detect anomalies by correlating apparatus and activity information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pattern recognition means is used to estimate apparatus states based on current data, then apparatus state detection is achieved, but learning all combinations of operating states requires enormous time and effort
Solution Approach 1:
The patent segments the apparatus state detection task by separating the learning phase (creating reference data for specific apparatus types) from the detection phase (matching current data against stored references). This allows the system to learn only characteristic patterns of known apparatus types rather than all possible apparatus state combinations, significantly reducing learning time while maintaining detection accuracy.
Solution Approach 2:
The patent applies preliminary action by pre-storing reference data for various apparatus types and their operating states before actual detection occurs. The feature quantity calculation means and apparatus state determination means are prepared in advance with reference patterns, enabling rapid matching during detection without requiring real-time learning of all possible states.
2Device complexity
If anomaly detection is based on absence of person's movements, then detection system structure is simplified, but detection errors occur when person is sleeping
Solution Approach 1:
The patent merges multiple detection approaches by combining presence detection (detecting whether a person is in the residence) with anomaly detection (detecting unusual patterns of apparatus operation). This integration allows the system to maintain simple structure while improving reliability through cross-validation of multiple data sources, distinguishing between normal sleep patterns and actual anomalies.
Solution Approach 2:
The patent implements feedback mechanisms where the anomaly detection means continuously monitors apparatus states and compares them against learned patterns and presence information. When inconsistencies are detected (e.g., apparatus operation without presence or unusual patterns during sleep hours), the system generates alerts, creating a feedback loop that improves detection accuracy over time.
Data Source
AI summary
An apparatus state detector and its associates are provided to save time and effort for learning combinations of operation states of all apparatuses at home and commit fewer estimation errors even when an unknown apparatus starts operating. The apparatus state detector includes measuring means that measures a physical quantity of an environment in which an apparatus is placed, feature-quantity calculation means that calculates a feature quantity of the measured value measured by the measuring means, storage means that stores in advance the feature quantity of each apparatus and an apparatus state associated with the feature quantity in a reference-apparatus entry dictionary, and apparatus-state detection means that searches the reference-apparatus entry dictionary for a feature quantity by using a feature quantity calculated by the feature-quantity calculation means as a search key and detects an apparatus state based on the apparatus state associated with the retrieved feature quantity.