Artificial hands can now be controlled through an app or by using sensors placed in the muscles of the forearm, according to research conducted at the Technical University of Munich (TUM). A breakthrough in understanding muscle activity patterns in the forearm has led to a more natural and intuitive control of artificial limbs. This innovative approach involves a network of 128 sensors and utilizes artificial intelligence techniques.
Over the years, technological advancements have resulted in the development of advanced artificial hands that offer greater functionality. These prostheses allow amputees to regain some movements, such as independent finger movements and wrist rotation. Control of these movements can be achieved through a smartphone app or by using muscle signals from the forearm, detected by two sensors.
For example, when the wrist flexor muscles are activated, the fingers close together to grip an object like a pen. Conversely, contracting the wrist extensor muscles causes the fingers to open, releasing the object. By simultaneously activating both flexor and extensor muscle groups, different finger movements can be controlled.
However, patients typically need to learn these movements as part of their rehabilitation process. To make controlling advanced hand prostheses more intuitive, Professor Cristina Piazza and her research team at TUM have employed artificial intelligence and the "synergy principle" with the help of 128 forearm sensors.
The synergy principle is based on the observation that the brain activates a pool of muscle cells, even in the forearm, when performing repetitive movements. When humans use their hands to grasp an object, such as a ball, their fingers move in a synchronized manner and adapt to the shape of the object upon contact. By leveraging this principle, the researchers have developed new learning algorithms to design and control artificial hands.
To achieve fluid and seamless movements, multiple steps are involved in controlling an artificial hand to grasp an object, such as a pen. The patient first orients the artificial hand to the grasping location, then slowly moves the fingers together before finally grabbing the pen. Through machine learning, the team aims to enhance control adaptability and improve the learning process, taking into account variations among individuals.
Experiments with the new approach, utilizing a greater number of sensors, have shown promising results, suggesting that conventional control methods can be enhanced with more advanced strategies. By using up to 64 sensors on the inside and outside of the forearm, the researchers can capture information from different muscle groups and determine which muscle activations are responsible for specific hand movements. This understanding is crucial for achieving intuitive movements.
The research primarily focuses on wrist and hand movements, and it has been observed that eight out of ten people prefer the intuitive way of moving the wrist and hand. This method is also more efficient. However, two out of ten individuals learn to handle the less intuitive way and ultimately become even more precise.
Dr. Patricia Capsi Morales, a senior scientist in Professor Piazza's team, explains that the research aims to investigate the learning effect and find the most suitable solution for each patient. Professor Piazza emphasizes that the control of artificial hands encompasses individual mechanics and hand properties, patient training, interpretation, analysis, and machine learning, making it a multi-faceted challenge.
