办事指南

A helping hand

点击量:   时间:2019-03-07 11:08:12

By Duncan Graham-Rowe in Orlando IT CAN take months to learn how to use an electronic artificial limb, but it could soon take just a few minutes, thanks to a new control device that adapts to suit each individual rather than vice versa. Most amputees can move the muscles in the remaining part of their limb, and the electric signals produced as the muscles contract can be measured using electromyography (EMG) sensors. In this way muscular motions can be used to control prosthetic limbs in real time, but getting them to do what you want isn’t easy. Now researchers at the University of Tsukuba in Japan have developed an “evolvable hardware” (EHW) chip that is able to redesign itself to handle the signals from each individual in the most efficient way. The connections between its reconfigurable logic gates are determined by a sequence of data that can be changed at will. What makes it special is that the chip actually evolves new sequences in much the same way that living creatures evolve, using a genetic algorithm. The sequence of the data that describes the logic array can be treated like genetic information, and different sequences are combined or mutated to breed new offspring. The fitness of the offspring is then evaluated and the best of them bred together to produce more offspring. The process is repeated again and again for many cycles until the configuration of the chip has evolved to produce the most effective or “fittest” design to do the job. In this case, the evaluation is determined by how closely the output, or hand action, matches the expected output pattern for the particular muscular signal. The hand is trained by taking EMG samples while the patient performs one of six possible actions that the prosthetic hand is capable of, such as extend or grasp. When the action is performed the system is given a key to correspond to the intended output or action. For just 10 samples for each action the system scored an average of 81 per cent accuracy. What makes this approach so effective, says Isamu Kajitani, who presented his work to the conference last week, is that if a person’s muscle signals change as the muscles degenerate, as commonly happens, the chip can adapt to the muscle signals as they change. So far the chip has been tested only in simulations, but the researchers are now looking for people who want to get involved in clinical trials. The hand they are using is multifunctional, enabling the patient to move it in a number of different ways. “But we are now working on a hand with five fingers that can be moved independently,” he says. This should be strong enough and dextrous enough to open a can of soda,