Behavioral Modeling of Dynamic Nonlinear Distortions in 5G Wireless Transmitters Using Cascaded Augmented Real-Valued...

Area of Science:

• Wireless communication
• Signal processing
• Machine learning

Background:

• Neural networks are crucial for enhancing wireless communication performance in 5G and 6G.
• Modeling dynamic nonlinear distortions in wireless transmitters is essential for system efficiency.

Purpose of the Study:

• To propose a computationally efficient neural network model for behavioral modeling of dynamic nonlinear distortions.
• To reduce the complexity of neural network-based models without compromising performance in 5G systems.

Main Methods:

• A modular two-box neural network system, CAR-VANN (cascaded augmented real-valued artificial neural networks), was developed.
• The first box uses an augmented, memoryless input for low-complexity nonlinear distortion modeling.
• The second box, an ARVTDNN (augmented real-valued time-delay neural network), fine-tunes accuracy.

Main Results:

• The CAR-VANN model achieved performance comparable to the ARVTDNN.
• CAR-VANN demonstrated a significant reduction in parameters (35%–52%).
• The model offers a viable, computationally efficient alternative for 5G systems.

Conclusions:

• CAR-VANN provides an effective solution for modeling dynamic nonlinear distortions in wireless transmitters.
• This approach reduces computational complexity in 5G systems while maintaining high performance.
• The modular design offers flexibility and efficiency in neural network applications for wireless communication.