Nanoscale magnetic devices simulate neuron behavior to identify human audio signals
Brain neurons behave as non-linear oscillators, producing rhythmic activity and interactive processing information. From this inspiration to achieve high-density, low-energy neuro-computers requires a very large number of nano-scale linear oscillators. Let's take a look at the related content with Xiaobian.
A team of researchers from France, the United States, and Japan jointly developed a nanoscale magnetic device that mimics the behavior of neurons that can be used to identify human audio signals. The study was published in the journal Nature, and the team described how the device was built, how it worked, and how accurate the results were.
A neuromorphic computer is a computer device that works by simulating the way the human brain works. In such a system, it means that the equipment created by the researchers simulates the way neurons, synapses, etc. work. In this new attempt, the researchers built a nanoscale neuromorphic computer with 400 neurons arranged in an array on a computer chip and used to identify human audio signals. It is worth noting that such devices are usually analog rather than digital, and if they can be developed, it is expected to have some advantages that traditional computers do not have, such as reduced power consumption, training, and high data conversion speed.
Schematic diagram of a spin torque nano-oscillator in which gold is a non-magnetic gap between two magnetic layers, blue is magnetic, and silver is a magnetized nanolayer
Imitation neurons are represented by small three-layer columns. The three-layer structure has a non-magnetic spacing between the two layers of magnetic. The continuous current induces direct magnetization at the top of the neuron, and the secondary current causes the magnetization to oscillate in a stable manner. Using the array as a computer device, the researchers read a number aloud with a microphone, such as "1", which passes the sound to a digital processor and converts it into an electrical signal. The electrical signal is then passed to the neural analog chip, which the researchers call the reservoir. Another digital computer reads the oscillations of the neurons and analyzes them, then converts the results into a form that humans can recognize, such as displaying the number "1" on the display. Testing the equipment with a variety of sounds, the team found that its accuracy can reach 99.6%.
The device is obviously still in its infancy, but it is simply built for research, but it shows that the neuro-computer will have great application prospects in the future. Compared with traditional computers, the neuro-computer will provide us with more A new way to process information.
British Switch Socket W80 Series
Fused Socket,Remote Control Plug Socket,Modular Switch Socket,Chrome Switches And Sockets
Guangdong Jinli Electrical Appliance Co., Ltd , https://www.sokelectric.com