US20260149626
2026-05-28
Electricity
H04L25/0254
The patent application discusses a method and apparatus for estimating channel characteristics in wireless communication systems. It uses a convolutional neural network (CNN) model with dilated convolution blocks arranged in a parallel connection structure to estimate channel characteristics of a reception channel. This estimation is crucial for determining the delay spread level in the channel, which affects the overall performance of wireless communication systems.
Wireless communication systems like 4G and 5G employ technologies such as beamforming and massive MIMO to counteract path loss and enhance signal transmission. OFDM is used due to its robustness against interference and frequency-selective fading, which occurs when signals are attenuated differently across frequencies. Improving channel estimation in these systems is essential for enhancing cell coverage and throughput, and accurately estimating delay spread is a key factor in this process.
The proposed method involves obtaining a channel matrix in the frequency domain using least square (LS) channel estimation. This matrix is then used to configure input data for the CNN model. The CNN model, featuring multiple dilated convolution blocks with a parallel connection structure, estimates the channel characteristics. Based on these characteristics, the delay spread level of the reception channel is determined.
The CNN model undergoes iterative training using dilated convolution blocks until validation loss converges. These blocks have different dilation rates, affecting the receptive field distribution in the frequency domain. The method includes a training mode for iterative training and an operation mode for real-time estimation using the trained model. Additional training can be performed with actual channel data during operation.
The communication device, equipped with processors and memory, executes instructions to perform the described method. It can classify delay spread levels into multiple categories and output probability values indicating the likelihood of a channel characteristic belonging to a specific delay spread level. This approach enhances the device's ability to estimate channel characteristics and delay spread effectively, improving communication performance.