Estimation of channel MSE for ATSC 3.0 receiver and its applications

Liu, Yu-Sun; Huang, Chun-Hung; You, Shingchern D.

doi:10.1016/j.icte.2021.09.001

Cited by 3 publications

(3 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In SP 16_2, the scattered pilots are located on a uniform spacing of 32 subcarriers in each OFDM symbol, and the scattered pilots in even and odd OFDM symbols are separated by 16 subcarriers in the frequency domain. When performing ML-based channel estimation methods, a two-dimensional approach [25] is utilized to find CFRs at the pilot and pseudo-pilot subcarriers, which are then used as inputs for the ML models. The CFR at each pilot subcarrier is first estimated using the LS method.…”

Section: Pilot Pattern and Training Datamentioning

confidence: 99%

See 1 more Smart Citation

Machine Learning-Based Channel Estimation Techniques for ATSC 3.0

Liu,

You,

Lai

2024

Information

Self Cite

View full text Add to dashboard Cite

Channel estimation accuracy significantly affects the performance of orthogonal frequency-division multiplexing (OFDM) systems. In the literature, there are quite a few channel estimation methods. However, the performances of these methods deteriorate considerably when the wireless channels suffer from nonlinear distortions and interferences. Machine learning (ML) shows great potential for solving nonparametric problems. This paper proposes ML-based channel estimation methods for systems with comb-type pilot patterns and random pilot symbols, such as ATSC 3.0. We compare their performances with conventional channel estimations in ATSC 3.0 systems for linear and nonlinear channel models. We also evaluate the robustness of the ML-based methods against channel model mismatch and signal-to-noise ratio (SNR) mismatch. The results show that the ML-based channel estimations achieve good mean squared error (MSE) performance for linear and nonlinear channels if the channel statistics used for the training stage match those of the deployment stage. Otherwise, the ML estimation models may overfit the training channel, leading to poor deployment performance. Furthermore, the deep neural network (DNN)-based method does not outperform the linear channel estimation methods in nonlinear channels.

show abstract

Section: Pilot Pattern and Training Datamentioning

confidence: 99%

“…Note that 𝑛 is an even integer. When performing ML-based channel estimation methods, a two-dimensional approach [25] is utilized to find CFRs at the pilot and pseudo-pilot subcarriers, which are then used as inputs for the ML models. The CFR at each pilot subcarrier is first estimated using the LS method.…”

Section: Pilot Pattern and Training Datamentioning

confidence: 99%

Machine Learning-Based Channel Estimation Techniques for ATSC 3.0

Liu,

You,

Lai

2024

Information

Self Cite

View full text Add to dashboard Cite

show abstract

“…Patch sketches of the natural image of the manifold (red color) and super-finished patch were obtained with MSE (blue color) and GAN (orange color). The solution with MSE (Y. S. Liu et al, 2021) looks too smooth because it is a pixel-wise average of the possible solutions in the pixel space, while the GAN Pixel-wise loss function to handle inherent uncertainty and perform high-frequency lost recovery. The detail of texture is that minimizing MSE encourages finding the pixel average of a reasonable solution.…”

Section: Loss Functionmentioning

confidence: 99%

Super Resolution Generative Adversarial Networks for Image Supervise Learning

Lupitha

Santoso

2022

SinkrOn

View full text Add to dashboard Cite

The E-Tilang application system has been widely used to support modern traffic, whereas protocol roads in big cities in Indonesia are already widely used. In principle, the plate number detection tool uses image recognition for detection. Image number plates on vehicles cannot always be read clearly, this is what causes the detection method to be a problem if the image plate number is further processed. The method for processing the plate number image uses deep learning and computer vision methods. For the condition of the image plate number that is not clear, the process of improving the image resolution from low resolution to high resolution is carried out, by applying Generative Adversarial Networks. This method consists of two main parts, namely Generate and Discriminator. Generate serves to generate an image and the Discriminator here is to check the image, can the image plate number be read or not? So that if the image plate number cannot be read, then the process is carried out again to the Generator until it is received by the Discriminator to be read. The process does not end here, the results will be carried out in the next process using Convolutional Neural Networks. Where the process is to detect the plate number image according to the classification of the plate number according to the region. The point is that an unclear image becomes clear by increasing the resolution from low resolution to high resolution so that it is easily read by the Convolutional Neural Network (CNN) algorithm so that the image is easily recognized by the CNN Algorithm. This becomes important in the CNN algorithm process because it gets the processed dataset. To produce a good model, preprocessing of the dataset is carried out. So that the model can detect the image well in terms of model performance.

show abstract