Touch Interface Force Estimation from Heatmaps Using Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing touch interfaces lack force-sensing functionality, which is costly and increases the size and complexity of computing systems, and existing methods for simulating force sensing using a single sensor are inaccurate in multi-touch scenarios.
Innovation Solution
A computing system utilizes a machine-learning model trained on data from multiple force sensors to simulate force-sensing functionality by analyzing touch heatmaps, allowing for accurate estimation of touch forces without physical force sensors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If physical force sensors are added to the touch interface, then force-sensing functionality is achieved, but system cost and device complexity increase
Solution Approach 1:
The patent creates a virtual copy of force-sensing functionality by training a machine learning model on data from physical force sensors. The model learns to predict force values from touch sensor data, effectively copying the sensing capability without requiring physical sensors in the deployed device. This resolves the contradiction by achieving force-sensing measurement precision through software simulation rather than hardware addition.
Solution Approach 2:
The patent replaces the mechanical/physical force sensing system with a computational approach. Instead of using physical force sensors that detect mechanical pressure, the system uses machine learning algorithms that process electrical signals from touch sensors to infer force values. This substitution eliminates the need for complex mechanical sensing components while maintaining force detection capability.
2Device complexity
If a single sensor is used to simulate force sensing, then system size is reduced, but measurement accuracy deteriorates in multi-touch scenarios
Solution Approach 1:
The patent segments the touch interface into multiple independent touch sensor elements that can individually detect touch events. By processing data from multiple segmented sensors simultaneously, the system can distinguish between multiple touch points and accurately estimate force for each touch independently. This segmentation enables multi-touch force sensing accuracy without requiring a single complex sensor.
Solution Approach 2:
The patent makes the touch sensor system multi-functional by enabling it to perform both touch detection and force sensing operations using the same hardware infrastructure. The machine learning model processes touch sensor data to extract multiple pieces of information including touch location, touch force, and multi-touch differentiation, making the single sensor system universally capable of multiple sensing functions.
Data Source
Figure 1A~1B
Figure 2~3
Figure 4
AI summary
Examples are disclosed herein relating to simulating force-sensing functionality for a touch interface using a machine-learning model that is trained based at least on training data generated by a training touch interface including a plurality of force sensors. In one example, a computing system (400) includes a touch interface (402) configured to output a touch heatmap (410) based at least on touch input detected by a plurality of touch sensors (404) of the touch interface (402). The computing system (400) is configured to execute a machine-learning model (418) that is configured to receive the touch heatmap (410), output a force estimation (420) of the touch input based at least on analyzing the touch heatmap (442), and execute a computing operation based at least on the force estimation (420). The machine-learning model (418) is trained based at least on training data (424) generated by a training touch interface including a plurality of force sensors.