Touch Recognition via Spatial-Temporal Feature Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing touch-enabled systems struggle to accurately distinguish between intended and unintended touches on touch-sensitive surfaces, leading to potential user experience issues due to unintended interactions.
Innovation Solution
A system that captures spatial features of touches over time intervals and compares them to stored sets of spatial features to determine whether a touch is intended, using machine-readable storage and processor-executable instructions to improve unintended touch rejection, with customization options based on user, application, and environment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional touch detection methods are used, then all touches are registered, but unintended touches cannot be distinguished from intended touches
Solution Approach 1:
The patent segments the touch detection process into multiple independent analysis dimensions: spatial feature extraction, temporal pattern analysis, and machine learning classification. By dividing the touch recognition problem into these separate components, the system can analyze each aspect independently and combine results to distinguish intended from unintended touches, thereby improving reliability without compromising ease of operation
Solution Approach 2:
The patent introduces temporal dimension by analyzing touch patterns over multiple time intervals and adding spatial feature dimensions through coordinate analysis. By transforming the touch detection from a single-point-in-time event to a multi-dimensional temporal-spatial analysis, the system gains the ability to distinguish touch intent without affecting user interaction naturalness
2Reliability
If touch detection is enhanced to distinguish intended vs unintended touches, then user experience improves, but system complexity increases
Solution Approach 1:
The patent implements self-service through automated machine learning models that learn touch patterns directly from device usage data without requiring manual configuration or calibration. The system automatically trains and updates its classification models based on observed user behavior, reducing the need for complex manual setup and maintenance while improving touch recognition reliability
Solution Approach 2:
The patent replaces traditional rule-based mechanical touch detection systems with data-driven machine learning algorithms. By substituting complex if-then logic rules with trained neural networks and classification models, the system achieves more accurate touch intent recognition while the learning capability gradually reduces the need for manual system configuration and adjustment
3Measurement precision
If machine learning models are trained with more data, then touch recognition accuracy improves, but processing time increases
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models offline with extensive touch pattern data before deployment. The models are pre-trained on diverse touch scenarios during manufacturing or initial setup, allowing them to make rapid real-time classifications during actual use without requiring extensive processing time for each touch event
Solution Approach 2:
The patent applies local quality by analyzing only the specific spatial and temporal features relevant to each touch event rather than processing all possible data. The system extracts only the necessary spatial coordinates, pressure values, and temporal patterns from each touch, enabling fast local decision-making at the touch detection point without global system-wide processing delays
Data Source
Figure 1~2
Figure 3~3B
AI summary
Examples relate to improving unintended touch rejection. In this manner, the examples disclosed herein enable recognizing a touch on a touch-sensitive surface, capturing a set of data related to the touch, wherein the set of data comprises a set of spatial features relating to a shape of the touch over a set of time intervals, and determining whether the recognized touch was intended based on a comparison of a first shape of the touch at a first time interval of the set of time intervals and a second shape of the touch at a second time interval of the set of time intervals.