MIT controls robot hand with ultrasound wristband
MIT has redefined the motion capture of human hands. Cameras, data gloves, or muscle signal acquisition are no longer needed for this.
(Image: MIT / Melanie Gonick)
Scientists at the Massachusetts Institute of Technology (MIT) have developed a control device that, when worn on the wrist, transmits the natural movements of a human hand to a robot hand. The researchers use small ultrasound sensors to capture hand movements. Previous techniques using cameras, data gloves, and muscle signal acquisition had proven too complicated or inaccurate.
The movements of a human hand result from the interplay of about 34 muscles, 27 joints, and more than 100 tendons and ligaments. To precisely capture and transmit these movements to a robot hand, various techniques have been developed, but they have some disadvantages. For example, motion capture via cameras is complex and susceptible to visual interference. Data gloves interfere with sensations and also restrict natural hand movements. A third technique, which captures hand movements via electrical muscle signals, has proven to be susceptible to interference from external signals and too insensitive to capture subtle movements.
Empfohlener redaktioneller Inhalt
Mit Ihrer Zustimmung wird hier ein externes YouTube-Video (Google Ireland Limited) geladen.
Ich bin damit einverstanden, dass mir externe Inhalte angezeigt werden. Damit können personenbezogene Daten an Drittplattformen (Google Ireland Limited) übermittelt werden. Mehr dazu in unserer Datenschutzerklärung.
Motion capture via ultrasound recordings
The MIT control device is about the size of a smartphone and is attached to the inside of the forearm above the wrist with a band. It uses miniaturized ultrasound sensors to capture the movements of the muscles, tendons, and ligaments of the wrist, as the researchers write in the study “Hand tracking using wearable wrist imaging,” which was published in Nature Electronics. Through the sensors, the researchers continuously take ultrasound recordings of the state of the wrist's musculoskeletal system. With the different positions of muscles, joints, tendons, and ligaments captured in this way, the researchers trained an artificial intelligence (AI) to derive the actual hand movements, for example in the form of gestures, in real-time and transmit them to a robot hand.
The researchers tested the system on eight test subjects. Hand movements were mostly recognized precisely. The system recognized the gestures for the 26 letters of American Sign Language without any problems. The system was also able to capture other subtle movement sequences, such as holding a pen, gripping a tennis ball, and pinch and zoom gestures, to control a robot hand and actions in virtual environments. The system also proved precise enough to capture playing the piano.
Videos by heise
In addition to controlling processes in virtual environments, the MIT scientists see applications for their technology primarily in the fine motor training of humanoid robots. Using the movement data of a human hand captured with the ultrasound wristband, robot hands can be precisely trained to perform difficult tasks. These include medical operations or complex manufacturing processes that traditionally require high levels of fine motor skills.
(olb)
