ML Correction of Sensor Readings for Thermal Interference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Temperature and humidity sensors in electronic devices face accuracy issues due to heat generated by system components, which affects the accuracy of ambient condition readings.
Innovation Solution
A machine learning model is trained to correct raw sensor values by accounting for heat sources, using input data such as power consumption, media playback state, and ambient light exposure to generate accurate ambient temperature and humidity values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If temperature and humidity sensors are mounted inside the electronic device to detect ambient conditions, then the device can obtain environmental data, but the heat generated by system components reduces the accuracy of sensor readings
Solution Approach 1:
The patent applies machine learning models to convert the harmful effect of internal heat on sensor readings into a beneficial correction mechanism. The system trains models using labeled data that captures the relationship between device state (power consumption, component activity) and sensor readings, then uses these models to calculate correction factors that compensate for thermal interference, transforming the problematic heat effect into a correctable parameter.
Solution Approach 2:
The system implements a feedback loop where sensor readings are continuously corrected based on device state information. The machine learning models process real-time data about power consumption, component activity, and environmental conditions to dynamically adjust correction factors, creating a closed-loop system that continuously improves measurement accuracy by feeding back correction information to the sensor readings.
2Measurement precision
If machine learning models are used to correct sensor values based on power consumption and device state, then measurement accuracy improves, but device complexity increases
Solution Approach 1:
The system applies preliminary action by pre-training machine learning models during a calibration phase using labeled data collected from multiple devices. These pre-trained models capture the thermal characteristics and interference patterns of specific device configurations. During actual operation, the pre-trained models are applied directly to correct sensor readings without requiring real-time training, thus improving accuracy while minimizing the computational complexity and power consumption during device operation.
Data Source
AI summary
A method is provided that includes reading a raw temperature value from a temperature sensor mounted in an electronic device and determining an amount of power applied to the electronic device. The method further includes generating, using a trained model, an ambient temperature value based on the raw temperature value and the determined amount of power, wherein the ambient temperature value represents a temperature outside of the electronic device.