New Algorithm Improves Control of Hand Protheses

Bielefeld University exhibiting at 2019 Hannover Trade Fair

From 1-5 April, Bielefeld University will be exhibiting at the 2019 Hannover Messe (Hannover Trade Fair) with the Research Institute for Cognition and Robotics (CoR Lab) and the Cluster of Ex-cellence Cognitive Interaction Technology (CITEC). The researchers will be presenting their platforms and applications for machine learning. One example of their work is a new method for quickly adjusting hand prostheses: The system enables flawless control of hand prostheses – even if the measuring electrodes have moved on the user’s skin. This system is one of four presentations by Bielefeld University, one of which belongs to a new start-up company founded by CITEC researchers. 

How does a hand prosthesis function properly even after the electrode cuff is readjusted? Researchers from Bielefeld University will demonstrate their system at the Hanover Trade Fair. Photo: Bielefeld University Interested visitors can try out the electrode cuff of the prosthetic hand system for themselves at Bielefeld University’s exhibition booth at the 2019 Hannover Messe (Hall 16, Booth A04). The exhibition booth is part of the joint exhibition stand of the Leading-Edge Cluster it’s Owl (Intelligent Technical Systems OstWestfalenLippe). “At the trade fair, we will be presenting advanced technologies and example applications for the efficient use of machine learning methods,” says Dr.-Ing. Sebastian Wrede of CoR-Lab, who is coordinating the participation in the trade fair. “These will be demonstrated following the value-added chain: from efficient hardware and software to integrated intelligent systems.”

Hand Prothesis Algorithm Compensates for Disturbances

Hand functioning can be partially restored with modern hand prostheses. Electrodes placed on the residual limb measure muscle signals,  an algorithm derives the desired hand movement from these signals, and a prosthesis then performs the movement. Such prosthetics, however, are prone to errors, especially if the electrodes move around on the skin. The Machine Learning research group, headed by Professor Dr. Barbara Hammer, has developed a system that compensates for errors caused by displaced electrodes. An machine learning algorithm adapts the control system from how it had been calibrated in the clinic to the new position of the electrode based on everyday use. What makes the system unique is that it gets by with very little data. “This also makes the new process appealing for industry,” says Sebastian Wrede. “Here, too, systems often have to make do with very little sample data.”

A Do-It-Yourself System for Object Recognition

Automatic object recognition is needed in many sectors of the economy, from the automotive industry to logistics. In order for a technical system to be able to recognize an object, it first has to recognize its defining features. The Cognitronics and Sensor Systems research group, headed by Prof. Dr.-Ing. Ulrich Rückert, has developed a mobile, cost-effective system called “TeachME,” which learns to recognize a new object and its characteristics in seconds flat. With the touch of a button, the system takes a picture of the object to be recognized and processes it with artificial neural networks that contain pre-trained models of objects. The system shows properties and other object data on its display. Small companies often do not have large computing capabilities and sometimes lack the expertise for machine learning. The new system is intuitive to operate and energy efficient – it can even be operated with a rechargeable battery.

Smart Mirror Gets by with Low Computing Power

The Smart Mirror, developed as part of an EU project, assists with schedules and provides tips for planning your day. The data is processed locally, ensuring that private data remains private. Photo: Bielefeld University As part of the EU project “Legato,” the Cognitronics and Sensor Systems research group has developed an intelligent mirror for smart homes. The project deals with energy-efficient data processing. Controlling in smart homes typically requires a great deal of computing power, and is mostly operated using cloud computing. Bielefeld’s smart mirror is intended to show how machine learning methods can be used onsite to save energy. The mirror recognizes its users and displays personalized information (such as bus schedules or current information about the home). It can be operated using both gestures and speech. And because the mirror processes data locally – not on external company servers – privacy is ensured.

Making Robots Accessible for Everyone

The tech start-up “R+” is dedicated to robotics and human-machine interaction. The team innovates products in areas of application ranging from customer support to healthcare. The goal is to relieve workers in their daily work routine by shifting repetitive tasks to robots, thereby freeing up more time for the human worker to do creative or caring tasks. The start-up provides its customers with a system that enables the average person to independently configure robots for their own individual needs. The system allows users to create solutions based on concepts from machine learning, machine vision, and edge computing (local data processing). With this system, the start-up aims to realize its vision of the robot as an service provider accepted in the heart of society.

Contact:
Dr.-Ing. Sebastian Wrede, Universität Bielefeld
Research Institute for Cognition and Robotics (CoR Lab)
Telefon: +49 521 106-5151 (secretary's office)
E-mail: swrede@cor-lab.uni-bielefeld.de