Myoelectric Signal Conversion for Teacher-Free Robot Motion Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for associating myoelectric potential with robot motion require user input of teaching signals, making effective learning difficult and laborious, and are challenging without using teacher data.
Innovation Solution
A conversion program that acquires feature vectors from biological signals, updates unit and class data based on energy attenuation and similarity, and outputs instruction signals to activate robots without relying on teacher data, using a computer to manage unit and class associations dynamically.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If teacher data is used to learn the association between biological signal and robot motion, then learning accuracy is improved, but user burden and system complexity increase
Solution Approach 1:
The system performs self-learning by automatically acquiring biological signals during robot operation and updating the association model without requiring external teaching data or user intervention. The learning process is embedded in the normal operation flow, allowing the system to improve its performance autonomously.
Solution Approach 2:
The system pre-acquires biological signals during robot operation before formal learning is needed, building up a database of signal-motion associations over time. This preliminary data collection enables the system to be ready for accurate recognition without requiring intensive teaching sessions when actually needed.
2Reliability
If teaching signals are required for learning, then learning effectiveness is improved, but operation time and user effort increase
Solution Approach 1:
The learning process continues continuously during robot operation without interrupting the workflow. Biological signals are acquired and processed in real-time, allowing the system to learn effectively while the robot is being used, eliminating the need for separate teaching sessions.
Solution Approach 2:
The system accumulates biological signal data during normal operation as preliminary learning material, so that when learning is actually needed, the system already has sufficient data to establish accurate associations without requiring additional time for teaching.
3Measurement precision
If manual association input is required, then learning precision is improved, but ease of operation deteriorates
Solution Approach 1:
The system automatically performs the association learning process by comparing acquired biological signals with robot motion data, eliminating the need for manual input. The algorithm independently identifies patterns and establishes associations between signal features and robot actions.
Solution Approach 2:
The manual mechanical process of teaching signals is replaced with an automated computational system that processes biological signals and learns associations through algorithmic analysis, substituting human effort with automated information processing.
4Adaptability or versatility
If the system adapts to user changes in biological signal patterns, then adaptability is improved, but system complexity increases
Solution Approach 1:
The system dynamically updates the association model as new biological signal patterns are acquired during operation. The learning model is not fixed but continuously adapts to changes in user behavior, physiology, or environment by incorporating new data into the existing framework.
Solution Approach 2:
The same signal acquisition and processing system serves multiple functions: both controlling the robot in real-time and simultaneously learning/adapting to user patterns. This multi-functionality allows adaptability without requiring separate dedicated teaching hardware or systems.
Data Source
AI summary
A conversion device includes: a unit attribute update unit that specifies a unit associated with a representative value approximating a feature vector for a biological signal, and updates a representative value and energy of the specified unit; a unit update unit that, when a plurality of feature vectors outside a predetermined range of a representative value of a unit of the unit data are acquired within a predetermined time, adds data for a new unit identifier to unit data; a class update unit that, when there is no class including a unit having energy within a predetermined range of energy of the new unit, adds data for a new class identifier to class data; and a motion update unit that updates motion data by associating, with an identifier of a class acquired after updating the class data, an identifier of a motion corresponding to the class.