Behavior identification device, air conditioner, and robot control device
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional behavior identification devices require pre-definition of component behaviors and configuration of identification devices, limiting their ability to identify complex behaviors in daily life, as they cannot flexibly define all possible combinations of component behaviors.
Innovation Solution
A behavior identification device that calculates sensor value distributions and uses Latent Dirichlet Allocation (LDA) to estimate basic distributions and component ratios, allowing for the identification of behaviors without pre-defining component behaviors, by treating sensor value distributions as vectors and comparing them to stored ratios.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If component behaviors are defined in advance and identification devices are configured for each component behavior, then behavior identification can be performed using structured component combinations, but the ability to identify complex behaviors in daily life is limited
Solution Approach 1:
The system performs self-learning by automatically acquiring sensor value distributions for various behaviors and deriving basic distributions and component ratios without requiring manual pre-definition. The behavior identification device serves itself by learning from collected data, eliminating the need for designers to pre-configure component behaviors while maintaining reliable identification accuracy
Solution Approach 2:
The system changes from a fixed pre-defined parameter approach to a dynamic learning approach where basic distributions and component ratios are automatically derived from sensor data. By treating sensor value distributions as vectors and applying mathematical transformations, the system adapts parameters based on actual behavior data rather than static pre-configuration
2Ease of manufacture
If a designer specifically defines component behaviors constituting a behavior in advance, then the identification device can be configured for those specific behaviors, but behaviors with complicated combinations of component behaviors cannot be identified
Solution Approach 1:
The system eliminates the need for designer intervention in defining component behaviors by implementing self-learning functionality. The behavior identification device automatically acquires sensor data, calculates sensor value distributions, and derives basic distributions and component ratios on its own, making the configuration process unnecessary while expanding behavior identification capabilities
Solution Approach 2:
The system performs preliminary data collection and analysis by acquiring sensor value distributions for various behaviors before actual identification is needed. By pre-calculating basic distributions and component ratios from collected data and storing them in databases, the system prepares identification capabilities in advance without requiring manual configuration when behaviors are encountered
Data Source
AI summary
The present invention provides a behavior identification device that can identify various behaviors without specifically defining a component constituting a behavior in advance. The behavior identification device comprising:a sensor-value obtaining unit (10) that obtains the sensor value and calculates a sensor value distribution that is a distribution of the sensor value measured within a predetermined time;a component database (42) that stores therein a set of basic distributions that are basic components constituting the sensor value distribution;a ratio calculating unit (21) that calculates a first component ratio that is a ratio of each of the basic distributions included in the sensor value distribution;a component ratio database (43) that stores therein a second component ratio that is the ratio determined in association with a behavior to be identified; andan identification unit (22) that compares the first component ratio to the second component ratio to identify the behavior,wherein the basic distribution is calculated as a sensor value distribution that is a base when each sensor value distribution is assumed to be a vector based on a set of the sensor value distributions obtained in advance for each of a plurality of types of the behavior.