Touch Recognition Models for Accurate Multi-Type Input Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electronic devices struggle to accurately identify various types of touch inputs, such as force, normal, and drag touches, due to variations in user interaction patterns, which affects the device's ability to provide an enhanced user experience.
Innovation Solution
The electronic device employs a method and system that utilizes a touch recognition model set comprising multiple models trained for different partial time periods of a touch input, applying weights to probability values to accurately identify touch types by combining them, and includes a processor to execute these models and manage training data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple touch recognition models are used for different time periods, then touch type identification accuracy is improved, but device complexity increases
Solution Approach 1:
The touch recognition system is segmented into multiple specialized models, each trained on specific time periods (e.g., early contact phase, sustained pressure phase, release phase). This segmentation allows each model to focus on particular touch characteristics, improving overall identification accuracy while maintaining manageable complexity through modular architecture
Solution Approach 2:
The system dynamically selects and applies different touch recognition models based on the current time period of the touch input being analyzed. This dynamic approach allows the system to adapt to varying touch characteristics at different stages, improving precision without requiring a single overly complex model
2Measurement precision
If touch recognition models are trained on different time periods, then identification of user intent is improved, but training data requirements increase
Solution Approach 1:
Training data is segmented into distinct time period subsets, with each subset used to train a specific touch recognition model. This segmentation allows the system to use smaller, more focused training datasets for each model rather than requiring one massive dataset, reducing overall data requirements while improving针对性 training effectiveness
Solution Approach 2:
Each touch recognition model is trained with high-quality, locally-optimized training data specific to its time period, rather than using generic or uniformly distributed data. This local quality approach ensures that each model receives the most relevant training examples for its specific function, improving intent identification accuracy with efficient data utilization
3Measurement precision
If weighted combination of probability values is used, then touch type recognition accuracy is improved, but computational complexity increases
Solution Approach 1:
The system uses weighted combinations of probability values from different time period models, where the weights are predetermined parameters optimized to reflect the importance of each time period. This approach improves recognition accuracy through sophisticated probability integration while avoiding the need for complex real-time calculations, balancing precision with computational efficiency
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Provided is a method of identifying a touch type of a user touch input with respect to an electronic device, the method including: obtaining touch data from a touch input received from a user; determining a touch recognition model set consisting of touch recognition models to be used to identify a touch type of the touch input of the user from among a plurality of touch recognition models corresponding to a plurality of partial time periods included in a time in which the touch input is maintained; obtaining touch type probability values with respect to the touch input of the user by applying the touch data to the touch recognition models included in the touch recognition model set; and identifying a touch type of the touch input, based on the obtained touch type probability values.