Bed with skin temperature estimations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing bed systems lack accurate methods for estimating skin temperature and adjusting comfort settings based on user movement and biometric data, leading to suboptimal sleep quality and comfort.
Innovation Solution
A bed system equipped with a linear array of temperature sensors and a two-part model using a gradient boosted decision tree and random forest algorithm to estimate skin temperature, combined with pressure sensors to detect user movement and biometrics, allowing for dynamic adjustment of firmness, temperature, and other comfort settings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If a linear array of temperature sensors spanning the entire bed is used to cover all possible user positions, then the measurement coverage is improved, but the complexity of selecting and processing data from the sensors increases
Solution Approach 1:
The bed is divided into multiple zones along its length, with temperature sensors distributed across these zones. The system selectively activates and processes data from only the sensors in the zone where a user is detected, rather than processing data from all sensors simultaneously. This segmentation approach maintains comprehensive coverage while reducing processing complexity.
Solution Approach 2:
The system uses more sensors than strictly necessary (excessive action) by deploying a linear array across the entire bed, but then applies partial action by selectively processing data from only the relevant subset of sensors in the active zone. This resolves the contradiction between having sufficient coverage and managing processing complexity.
2Measurement precision
If temperature sensors are positioned close to the user's body to capture accurate readings, then the measurement precision is improved, but the system cannot directly measure distal body temperatures without additional processing
Solution Approach 1:
A two-stage machine learning model serves as an intermediary between the proximal temperature sensor readings and the desired distal temperature estimates. The gradient boosted decision tree model first generates candidate distal temperature values from proximal sensor data, then the random forest model refines these estimates and reduces the impact of noisy or anomalous readings, effectively bridging the information gap.
Solution Approach 2:
The system replaces direct physical measurement of distal body temperatures with a computational modeling approach. Instead of placing sensors at various body locations, the system uses machine learning models to infer distal temperatures from proximal sensor readings, substituting mechanical measurement with intelligent algorithms.
3Ease of operation
If the bed system continuously monitors and adjusts comfort settings based on real-time user data, then the user comfort is improved, but the energy consumption and system complexity increase
Solution Approach 1:
The system performs temperature monitoring and comfort adjustments periodically rather than continuously. It monitors user presence and activates temperature estimation and adjustment functions when a user is detected in the bed, then reduces processing when no user is present, thereby reducing energy consumption while maintaining user comfort.
Solution Approach 2:
The bed system automatically monitors user temperature and adjusts comfort settings without requiring manual user input or intervention. The machine learning models and control systems operate autonomously to maintain optimal comfort conditions, reducing the need for active user management while conserving energy through intelligent, event-driven operation.
Data Source
AI summary
A bed has a mattress. A plurality of temperature sensors are each temperature sensor configured to: sense temperature in a particular location of the bed; and transmit, to a computer system, a datastream of the sensed temperature. A computing system is configured to: receive each of the datastreams of the sensed temperature; select a subset of the datastreams as selected datastreams; aggregate the selected datastreams into time-bound windows of temperature data; select a subset of the time-bound windows of temperature as selected temperature values; access a machine-learning temperature classifier; provide, to the machine-learning temperature classifier, the selected temperature values; and receive, from the machine-learning temperature classifier, the distal wrist temperature for the user.