Myoelectric Signal Conversion for Teacherless Motion Learning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods require users to manually input motion signals for electric hands, which can lead to difficulties in associating myoelectric potentials with hand motions and hinder effective learning.

Innovation Solution

A conversion program and device that learns the association between myoelectric potentials and hand motions without using teacher data by updating energy and identifiers of feature vectors and classes based on user interactions, utilizing a diffusion equation to optimize the association.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If teacher data is used to learn the association between biological signals and robot motions, then learning accuracy can be improved, but user burden and system complexity increase

Engineering Contradiction:
Improvelearning accuracyVSAvoiduser burden
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The system performs self-learning by automatically acquiring biological signals during normal robot operation and updating the association model without requiring user intervention for data collection. The control device autonomously identifies feature vectors from biological signals and updates the learning model in real-time, eliminating the need for users to manually provide teaching data.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system changes the parameter representation by using energy values derived from biological signal feature vectors as learning parameters. Instead of requiring labeled teaching data, the system extracts energy information from the biological signals themselves and uses this as the basis for learning the association between biological signals and robot motions.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If manual teaching input is required for learning, then learning precision can be improved, but ease of operation deteriorates

Engineering Contradiction:
Improvelearning precisionVSAvoidease of operation
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The system performs self-learning by automatically acquiring biological signals during normal robot operation and updating the association model without requiring user intervention for data collection. The control device autonomously identifies feature vectors from biological signals and updates the learning model in real-time, eliminating the need for users to manually provide teaching data.

Inventive Principle:
Principle #25Self-service

3Stability of the object's composition

If the association between biological signal and robot motion is fixed, then system stability is improved, but adaptability deteriorates

Engineering Contradiction:
Improvesystem stabilityVSAvoidadaptability
Core Design Contradiction:
Stability of the object's compositionVSAdaptability or versatility

Solution Approach 1:

The system dynamically updates the association between biological signals and robot motions by continuously learning from new biological signal data. The control device periodically acquires new feature vectors, calculates their energy values, and updates the learning model to reflect changing user intentions and muscle patterns, allowing the system to adapt while maintaining operational stability through controlled learning updates.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP4089507B1Conversion program and conversion device
Publication Date: 2025.11.19 UNIVERSITY OF ELECTRO-COMMUNICATIONS
  • EP4089507B1 patent drawingFigure 1
  • EP4089507B1 patent drawingFigure 2
  • EP4089507B1 patent drawingFigure 3

AI summary

A conversion device 1 includes: an acquisition unit 21 configured to sequentially acquire feature vectors for a biological signal; a unit attribute update unit 22 configured to specify a unit associated with a representative value approximating a new feature vector, and update a representative value and energy of the specified unit; a unit update unit 23 configured to, 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, add data for a new unit identifier to unit data 12; a class update unit 24 configured to, when there is no class including a unit having energy within a predetermined range of energy of the new unit, add data for a new class identifier to class data 13; and a motion update unit 25 configured to update motion data 14 by associating, with an identifier of a class acquired after updating the class data 13, an identifier of a motion corresponding to the class.