Scholarlink

In recent days, the wireless communication system requires efficient channel management for the effective data transmission. However, the increasing demand and the limited resources are raising significant issues in network performance and capacity. The existing models majorly suffered with interpretability and handling long-term dependencies. In this research, the graph attention with temporal convolutional network (GAT-TCN) is proposed for effective channel management. Initially, the input signals are considered from channel state information (CSI) dataset and further preprocessed with compressed sensing (CS) denoising technique. The CS denoising technique remove noise in order to improve interpretability by enhancing the signal quality. After that, the denoised signals are further processed with Stockwell transform (ST) to extract time–frequency domain features. Finally, the proposed GAT-TCN is utilized for the prediction of sparse signals, and the channels are managed with the help of resource allocation and reallocation techniques. From the results, the proposed GAT-TCN attained better results in terms of bit error rate (BER) of 77% and mean square error (MSE) of 84% when compared to existing auto-encoder-bidirectional long short-term memory (AE-BiLSTM).

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Sparse Signal Processing for Efficient Channel Management in Wireless Network Using Graph Attention with Temporal Convolutional Networks

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Shivaranjani Shirabadagi,
  • Ensteih Silvia,
  • P. N. Siva Jyothi,
  • P. R. Ajitha,
  • Supriya

摘要

In recent days, the wireless communication system requires efficient channel management for the effective data transmission. However, the increasing demand and the limited resources are raising significant issues in network performance and capacity. The existing models majorly suffered with interpretability and handling long-term dependencies. In this research, the graph attention with temporal convolutional network (GAT-TCN) is proposed for effective channel management. Initially, the input signals are considered from channel state information (CSI) dataset and further preprocessed with compressed sensing (CS) denoising technique. The CS denoising technique remove noise in order to improve interpretability by enhancing the signal quality. After that, the denoised signals are further processed with Stockwell transform (ST) to extract time–frequency domain features. Finally, the proposed GAT-TCN is utilized for the prediction of sparse signals, and the channels are managed with the help of resource allocation and reallocation techniques. From the results, the proposed GAT-TCN attained better results in terms of bit error rate (BER) of 77% and mean square error (MSE) of 84% when compared to existing auto-encoder-bidirectional long short-term memory (AE-BiLSTM).