Mobile Device Precipitation Detection via Sensor Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In regions lacking conventional precipitation sensors and infrastructure, existing methods for quantitative precipitation estimation are unreliable due to limited data availability and interpolation errors, especially in areas with limited communication tower density and high humidity or frozen precipitation conditions.
Innovation Solution
A method utilizing machine learning models trained with historical data from mobile devices to detect precipitation by analyzing signal strength and location, leveraging both cellular and WiFi communication networks, and incorporating user input and image classification for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional precipitation sensors and infrastructure are used, then measurement precision is improved, but device complexity and infrastructure requirements increase
Solution Approach 1:
The patent enables mobile devices to perform precipitation detection independently using their existing sensors (accelerometers, gyroscopes, barometers, microphones, cameras) without requiring external precipitation sensing infrastructure. The machine learning model processes data from these self-contained device sensors to detect precipitation, allowing the device to serve its own detection needs.
Solution Approach 2:
The patent replicates precipitation detection functionality using data from mobile device sensors instead of dedicated precipitation sensors. By copying the detection function to ubiquitous mobile devices and using their existing sensor data (motion, pressure, audio, visual), the system achieves precipitation detection without complex infrastructure.
2Area of stationary object
If radar and satellite data are used for quantitative precipitation estimation, then coverage area increases, but measurement precision deteriorates in data-poor regions
Solution Approach 1:
Mobile devices independently perform precipitation detection using local sensor data and machine learning models, without relying on external radar or satellite infrastructure. Each device serves its own detection needs using its onboard sensors, enabling accurate detection in data-poor regions where remote sensing is unavailable or inaccurate.
Solution Approach 2:
The system changes the detection parameters from remote sensing (radar/satellite) to local device sensor measurements (acceleration, pressure, audio, visual data). This parameter transformation enables precipitation detection to work effectively in regions where traditional remote sensing infrastructure is limited or nonexistent.
3Measurement precision
If machine learning models are trained with historical data from mobile devices, then precipitation detection accuracy in data-poor areas improves, but data processing complexity increases
Solution Approach 1:
The machine learning models are trained in advance using historical data from mobile devices and quantitative precipitation estimation data. This preliminary training phase creates ready-to-use models that can then perform rapid precipitation detection without requiring complex real-time processing during actual detection events.
Solution Approach 2:
The machine learning model acts as an intermediary that processes and interprets raw sensor data from mobile devices. Instead of requiring complex real-time analysis of multiple sensor inputs, the pre-trained model serves as a mediator that translates sensor data into precipitation detection results, simplifying the processing complexity.
Data Source
AI summary
In an approach for precipitation detection, a processor trains a first machine learning model for detecting precipitation in a region using a first training set of data including a plurality of historical data from a plurality of mobile devices collected in the region and a plurality of quantitative precipitation estimation data. A processor trains a second machine learning model for detecting a location of a mobile device in the region using a second training set of data including both historical indoor and outdoor data from the plurality of mobile devices collected. A processor receives a current data from the mobile device. A processor determines whether the mobile device is located indoor or outdoor based on the current data. A processor compares the current data against a threshold set in the first machine learning model to indicate precipitation. A processor determines whether the current data exceeds the threshold.