| Summary: | Comparing full-duplex systems to time-division duplex and frequency-division duplex systems, the efficiency of spectrum use can be theoretically doubled. It is one of the development directions of next-generation wireless communication technologies. One of the critical issues to be addressed in full-duplex systems is the self-interference generated by their own transmitted signals, which can affect the correct demodulation of useful signals.
This report offers a digital deep learning-based self-interference cancellation technique for full-duplex networks in response to the traditional digital-domain self-interference cancellation techniques' poor cancellation ability.
The primary sources of interference in the self-interference channel are investigated, and a theoretical model of the self-interference channel in a full-duplex system is developed. The deep learning-based self-interference cancellation model is strengthened by an algorithm that is created to extract features from the reference signal that can remove outlier feature values in the data.
Additionally, a deep learning-based digital-domain self-interference cancellation technology for full-duplex is provided. This technology is based on the primary nonlinear sources in the self-interference channel model. The simulation's findings show that linear cancellation provides around 37.86 dB. For nonlinear components, the LMS adaptive filtering algorithm offers roughly 3.88 dB of suppression; the deep learning network model's suppression capability is 8.08 dB, which puts it nearly at the receiver noise floor.
We also analyze the impact of different network complexity on the self-interference cancellation capability. Besides, the quantization relationship between the loss function value of the deep learning network and the self-interference cancellation capability were simulated and analyzed.
The report examined the self-interference cancellation method for full-duplex networks and performed simulation analysis on the method. It offers a fresh approach and new lines of inquiry for enhancing the effectiveness of self-interference suppression in the digital domain of full-duplex systems.
Key words: Self-interference cancellation, Deep learning, Full-duplex network
|
|---|