Deep learning enhances Reedberg multi-frequency microwave recognition

The team of academician Guo Guangcan of the University of Science and Technology of China has made new progress in multi-frequency microwave sensing. Professors Paulson and Ding Dongsheng’s research group used artificial intelligence methods to achieve precision detection based on Reedberg atomic multi-frequency microwaves. The results were published in Nature Communications on April 14.

Decode the results for machine learning. (a-c) The recovery result of the deep learning model on the transmitted signal for different training times. png

The figure shows the machine learning decoding result. (a-c) For the recovery results of the deep learning model on the transmitted signal at different training times, courtesy of the University of Science and Technology of China

The Reedberg atom has a large electric dipole moment and can produce a strong response to weak electric fields, so it is favored as a very promising microwave measurement system. However, there are still many scientific problems in the field of microwave measurement based on Reedberg atoms that need to be solved, and multi-frequency microwave reception is one of the difficulties: this is because multi-frequency microwaves cause complex interference patterns in atoms, which seriously interfere with signal reception and recognition.

In recent years, the team of Splausson and Ding Dongsheng have made important progress in using the Reedberg atomic system to focus on quantum simulation and quantum precision measurement scientific research. In this study, based on the room temperature rubidium atomic system, the team used the Reedberg atom as a microwave antenna and modem to successfully detect the multi-frequency microwave field of phase modulation (binary phase shift keying signal with frequency division multiplexing, a signal transmission method widely used in digital communication) through electromagnetic induction transparency effect, and then analyzed the received modulated signal through the deep learning neural network, realized the high-fidelity demodulation of multi-frequency microwave signals, and further tested the high robustness of the experimental scheme for microwave noise.

This work effectively decodes a frequency-division multiplexed phase-shift keyed signal of a noisy QR code with an accuracy rate of 99.32%. The results show that the Rydberg microwave receiver based on deep learning enhancement can allow direct decoding of 20 frequency division multiplexed signals at a time, without the need for multiple band-pass filters and other complex circuits.

The innovation of this work is that it proposes and implements a scheme for effectively detecting multi-frequency microwave electric fields without solving the main equation, taking advantage of the sensitivity advantages of The Reedberg atoms while also reducing the effects of noise. This work organically combines atomic sensing with deep learning, providing an important reference for the cross-binding of precision measurements with neural networks. In addition, the results can be applied to detect multiple targets at the same time.

The reviewers believe that “the results presented by this work are very useful to other researchers in the field of atomic and molecular photophysics, as it shows the future application of deep learning in quantum-enhanced sensing of atomic systems.” (Source: China Science Daily Wang Min)

Related paper information:

Source link

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button