Finger Handwriting Recognition with Contextual Word Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing text input systems face challenges in recognizing handwritten characters, particularly when using fingers, due to variance in character appearance and the need for dedicated input areas, which complicates user interaction on mobile devices with limited computing power and screen size constraints.
Innovation Solution
A text input system that combines handwriting recognition with word prediction, allowing multiple character candidates to be recognized and filtered based on context, enabling users to select intended words without writing entire characters, and utilizing a neural network for improved character recognition and a touch-sensitive interface for finger input.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If dedicated input areas are introduced for different character subsets, then character recognition accuracy is improved, but device complexity and user learning burden increase
Solution Approach 1:
The patent applies universality by using a single unified input field that accepts all character types (uppercase, lowercase, digits, punctuation) without requiring separate dedicated areas. The neural network recognizer is trained to handle all character variants within this single field, eliminating the need for multiple specialized input zones while maintaining recognition accuracy.
Solution Approach 2:
The patent changes the approach from spatial separation (different areas for different characters) to parameter-based recognition (using neural network parameters to distinguish character types). The system uses pixel density, stroke patterns, and contextual parameters rather than positional constraints to identify character subsets, allowing all characters to be recognized within one unified field.
2Ease of operation
If finger-based handwriting input is supported, then ease of operation is improved, but character recognition accuracy deteriorates due to larger stroke variance
Solution Approach 1:
The patent applies dynamics by training the neural network to adapt to varying stroke characteristics inherent in finger writing. The system dynamically adjusts to different writing speeds, pressures, and stroke sizes that occur with finger input, rather than requiring fixed, rigid character patterns. This allows natural finger handwriting variations to be accommodated while maintaining recognition accuracy.
Solution Approach 2:
The patent uses preliminary action by pre-training the neural network with extensive finger handwriting data before actual use. The system is prepared in advance to recognize the specific characteristics of finger-written characters, including larger stroke variations and different pressure patterns, so that accurate recognition occurs automatically during normal operation without requiring real-time adjustment.
3Productivity
If word prediction is implemented, then text input efficiency is improved, but computing power requirements increase
Solution Approach 1:
The patent applies partial action by implementing word prediction that generates and displays only a limited number of candidate words (e.g., top 5-10 predictions) rather than processing all possible word combinations. The system performs partial word completion matching and filters candidates based on context, reducing the computational burden while still providing efficient text input through prediction of the most likely next words.
4Measurement precision
If multiple character candidates are recognized, then recognition accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent uses feedback by implementing an iterative recognition process where multiple character candidates are generated, then filtered and refined based on contextual feedback from the input field and previously recognized characters. The system provides feedback to the user through candidate word displays, allowing selection that further refines future recognition, creating a feedback loop that improves accuracy without requiring overly complex processing at each individual step.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
The present invention relates to a text input system and method involving finger-based handwriting recognition and word prediction. A text input device (300) comprises: a text prediction component (310) for predicting a plurality of follow-up words based on a text context, the text prediction component (310) outputting a set of candidate words; a character handwriting recognition component (330) for recognizing a handwritten character candidate, the handwritten character candidate being determined based upon handwriting input received from a touch sensitive input field (340); a candidate word filtering component (350) for filtering the set of candidate words received from the text prediction component (310) based on the recognized handwritten character candidate; a word presentation component (360) for presenting candidate words from the filtered set of candidate words to a user of the device; and a word selection component (380) for receiving a user selection of a presented candidate word from the user.