Bed system for determining user biometrics during sleep based on load-cell signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing bed systems lack the capability to accurately and efficiently determine biometric information such as heart rate and respiration rate of users during sleep sessions without requiring additional sensors or processing units, and struggle with noise and signal quality issues.
Innovation Solution
The integration of load cells as force sensors in the bed system to collect and process load-cell signals using filtering techniques and machine learning models, such as deep neural networks, to determine biometric parameters like heart rate and respiration rate, even in the presence of noise and low signal quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If additional sensors or processing units are added to the bed system to determine biometric information, then measurement precision improves, but device complexity increases
Solution Approach 1:
The load cells in the bed system are made to serve dual purposes: supporting the bed structure and detecting biometric signals. By processing existing load-cell signals through filtering techniques and machine learning models, the system extracts heart rate and respiration rate information without adding dedicated biometric sensors, thus maintaining measurement precision while avoiding increased device complexity
Solution Approach 2:
The bed system's existing force sensors (load cells) are utilized to automatically detect and process biometric information. The system processes its own operational data (load-cell signals) to derive additional functionality, eliminating the need for separate sensing systems and reducing overall system complexity while maintaining accurate biometric measurement
2Measurement precision
If filtering techniques and machine learning models are applied to load-cell signals, then measurement precision improves, but use of energy increases
Solution Approach 1:
Signal filtering is applied as a preliminary step before machine learning processing to remove noise and enhance relevant biometric patterns in the load-cell signals. This preprocessing reduces the complexity of subsequent machine learning analysis, allowing accurate heart rate and respiration rate detection while minimizing the computational energy required for the more intensive machine learning operations
3Measurement precision
If load-cell signals are processed to determine biometric parameters, then measurement precision improves, but reliability decreases due to noise and low signal quality
Solution Approach 1:
The system converts the inherent noise and signal variations in load-cell measurements into useful biometric information. By applying machine learning models trained to recognize patterns in noisy signals, the system extracts reliable heart rate and respiration rate data even from low-quality load-cell signals, transforming what would be harmful interference into the detection mechanism itself
Solution Approach 2:
The machine learning models continuously analyze load-cell signals and provide feedback to refine signal processing parameters and improve detection accuracy over time. This adaptive feedback mechanism allows the system to maintain reliable biometric measurements despite varying signal quality conditions by learning from past performance and adjusting processing strategies
Data Source
AI summary
Disclosed are systems and techniques for determining biometrics of a user of a bed system based on force data. A bed system can include a support element having at least one leg, at least one force sensor of the at least one leg, the force sensor being configured to sense a force applied to the bed system or the leg, and a controller. The controller can receive at least one force data-stream from the at least one force sensor, the at least one force data-stream representing a force sensed by the force sensor, determine a biometric parameter of a user on the bed system at predetermined time intervals based on processing the at least one force data-stream, generate an aggregate biometric parameter of the user based on aggregating the biometric parameters for the predetermined time intervals, and return the aggregate biometric parameter of the user.