Handwritten Music Sign Recognition Using Vector Stroke Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing music sign recognition technologies require large storage capacity and complex programs to register specific signs, making them inefficient for recognizing unregistered characters and requiring precise input, while also increasing computational load for line component recognition.
Innovation Solution
A handheld music sign recognition device that captures vector information of input movements, derives stroke characteristics, and stores these for accurate recognition without pre-registration, using a touch information obtaining system, vector information creation, stroke information storage, and stroke characteristic derivation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If specific signs or figures are registered in advance for character conversion, then character recognition can be performed, but large storage capacity is required and the program becomes complicated
Solution Approach 1:
The patent segments the recognition process into two distinct phases: an off-line learning phase where the system adapts to the user's handwriting characteristics, and an on-line recognition phase where the segmented stroke data is matched against the learned model. This segmentation eliminates the need for pre-registering large databases of character samples, as the system dynamically creates a personalized recognition model through incremental learning from user input.
Solution Approach 2:
The system performs preliminary action by conducting off-line learning to establish a personalized handwriting model before actual recognition occurs. During this preliminary phase, the system collects and analyzes stroke data to create a reference model of the user's writing patterns, which is then used for efficient on-line recognition without requiring extensive pre-registered character databases.
2Adaptability or versatility
If large amounts of character data are stored for recognition, then more characters can be recognized, but the storage medium requires large storage capacity
Solution Approach 1:
The system implements self-service by automatically adapting to each user's handwriting style through off-line learning, creating a personalized recognition model without requiring pre-stored samples of that user's writing. The system serves itself by generating its own training data from the user's initial inputs and using this to build a customized recognition engine, eliminating the need for large universal character databases.
Solution Approach 2:
The patent changes the fundamental parameter from storing static character images to storing dynamic stroke trajectory data in vector format. This parameter change allows the system to represent characters through mathematical descriptions of stroke paths rather than pixel-based images, enabling efficient storage and flexible recognition of diverse characters without proportionally increasing storage requirements.
3Measurement precision
If line component recognition is performed sequentially with coordinate data processing, then line components can be extracted, but the amount of calculations becomes large
Solution Approach 1:
The patent replaces the mechanical coordinate-by-coordinate processing system with a vector-based mathematical model. Instead of sequentially analyzing individual coordinate points, the system represents strokes as continuous vector functions and uses parametric equations to describe stroke trajectories. This substitution transforms the recognition process from iterative numerical computation to analytical geometry operations, significantly reducing computational load while maintaining precision.
Data Source
AI summary
A handwritten music sign recognition device (10) has: a touch information obtaining part (21) which obtains, as position information, a position of an input means on a screen every time the input means moves, from when the input means touches the screen until when the input means is moved away therefrom; a vector information creation part (22) which creates attributes of a vector indicating a trace of movement of the input means as vector information based on the position information; and a stroke characteristic amount derivation part (23) which derives a characteristic amount of a stroke based on the vector information included in the stroke information.