Mobile Device Placement Detection via Sensor Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current technologies lack accurate methods to determine the placement of mobile electronic devices relative to a user, such as in a pocket, bag, or on a surface, which limits their ability to provide placement-dependent interactions and prevent unintended device operations.
Innovation Solution
The Placement Detector uses sensor-based probabilistic models trained with machine learning techniques to infer the placement of handheld devices by evaluating data from various sensors like accelerometers, capacitive arrays, and multi-spectral sensors, enabling the device to determine its position and initiate appropriate actions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If accelerometer and gyroscopic data are used to detect device placement, then the ability to detect hand-held or pocket placement is improved, but the ability to classify off-body locations (e.g., in a bag or on a desk) deteriorates
Solution Approach 1:
The patent combines data from multiple sensor types (accelerometer, gyroscope, light sensor, microphone, proximity sensor, capacitive sensor) to create a comprehensive sensor fusion approach that enables accurate classification of both on-body and off-body device placements, resolving the limitation of using single sensor types
Solution Approach 2:
The system implements a universal classification framework that handles multiple device placement scenarios (hand-held, pocket, bag, desk, table, etc.) using a single integrated sensor evaluation system, making the solution adaptable to various placement contexts beyond what individual sensors can achieve
2Measurement precision
If light sensor data is used to determine device placement, then the ability to distinguish pocket vs. out-of-pocket placement is improved, but the solution is limited to lab environments and specific placement scenarios
Solution Approach 1:
The patent merges light sensor data with multiple other sensor inputs (accelerometer, gyroscope, microphone, proximity sensor, capacitive sensor) to create a robust multi-sensor evaluation system that maintains accuracy across diverse environmental conditions, not just lab settings
3Measurement precision
If microphone and capacitive sensor data are used to infer device placement, then the ability to distinguish pocket vs. non-pocket placement is improved, but the solution requires multiple separate sensor types increasing device complexity
Solution Approach 1:
The system creates a universal sensor evaluation framework that efficiently processes data from multiple sensor types (accelerometer, gyroscope, light sensor, microphone, proximity sensor, capacitive sensor) through a unified probabilistic model, managing the complexity of multiple sensors through integrated processing
4Measurement precision
If sensor-based probabilistic models with machine learning are used to infer device placement, then the accuracy of placement inference is improved, but the computational resources and processing time required increase
Solution Approach 1:
The system implements partial action by selectively evaluating sensor data based on current context and device state, using probabilistic models to determine the most likely placement without always performing exhaustive analysis, thus reducing computational energy consumption while maintaining accuracy
Solution Approach 2:
The system uses feedback from sensor evaluations to refine placement inference over time, using machine learning techniques that adapt to user behavior patterns, reducing the need for continuous high-power processing while maintaining high accuracy through learned predictions
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This solution allows for accurate inference of device placement, enabling features like preventing pocket dialing, dynamic notification preferences, and location tracking, enhancing user-device interactions and improving user experience.
Implementation Method 1
uses a sensor-based probabilistic model trained with machine learning techniques to infer the placement of handheld devices by evaluating data from various sensors like accelerometers
Implementation Method 2
evaluating data from various sensors like accelerometers, capacitive arrays, and multi-spectral sensors
Data Source
AI summary
A “Placement Detector” enables handheld or mobile electronic devices such as phones, media players, tablets, etc., to infer their current position or placement. Placement inference is performed by evaluating one or more sensors associated with the device relative to one or more trained probabilistic models to infer device relative to a user. Example placement inferences include, but are not limited to, inferring whether the device is currently in a user's pocket, in a user's purse (or other carrying bag or backpack), in a closed area such as a drawer or box, in an open area such as on a table, indoors, outdoors, etc. These types of placement inferences facilitate a wide range of automated user-device interactions, including, but not limited to, placement-dependent notifications, placement-dependent responses to various inputs, prevention of inadvertent “pocket dialing,” prevention of inadvertent power cycling of devices, lost or misplaced device location assistance, etc.