Scholarlink

<p>The given paper introduces a novel framework to enhance signal to noise ratio (SNR) in Ultra-Wideband (UWB) transceivers implemented on Field-Programmable Gate Array (FPGA) platforms. The proposed framework makes use of Modified Moving Average (MMA) filter in conjunction with sparse autoencoder (SAE) network segment. The methodology involves oversampling, adaptive filtering, and decimation, validated through MATLAB simulations and hardware experiments. SAE has sparsity constraint which selects set of features for reconstruction of input signal thus leading to dimensionality reduction. Under the sparse constraint, the hidden layer of the SAE network can learn the sparse and concise feature representation of the original input data from the input layer. The results demonstrate a significant SNR improvement from 5.0 dB to 27.57 dB using a 25-tap filter with a 5× oversampling factor at a 250 MS/s sampling rate. The algorithm’s efficacy is further correlated with Effective Number of Bits (ENOB), adhering to the relationship SNR = 6.02×ENOB + 1.76 dB. This work bridges the gap between software-defined signal processing and hardware efficiency, offering a reconfigurable solution for next-generation wireless communication systems.</p>

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

A novel framework based on modified moving average (MMA) filter and sparse autoencoder (SAE) to enhance SNR ratio in Ultra-Wideband transceivers

  • 期刊论文

  • 发表日期：

    2025年12月6日

查看全文
  • V. Bindusree,
  • Madhavi Tatineni

摘要

The given paper introduces a novel framework to enhance signal to noise ratio (SNR) in Ultra-Wideband (UWB) transceivers implemented on Field-Programmable Gate Array (FPGA) platforms. The proposed framework makes use of Modified Moving Average (MMA) filter in conjunction with sparse autoencoder (SAE) network segment. The methodology involves oversampling, adaptive filtering, and decimation, validated through MATLAB simulations and hardware experiments. SAE has sparsity constraint which selects set of features for reconstruction of input signal thus leading to dimensionality reduction. Under the sparse constraint, the hidden layer of the SAE network can learn the sparse and concise feature representation of the original input data from the input layer. The results demonstrate a significant SNR improvement from 5.0 dB to 27.57 dB using a 25-tap filter with a 5× oversampling factor at a 250 MS/s sampling rate. The algorithm’s efficacy is further correlated with Effective Number of Bits (ENOB), adhering to the relationship SNR = 6.02×ENOB + 1.76 dB. This work bridges the gap between software-defined signal processing and hardware efficiency, offering a reconfigurable solution for next-generation wireless communication systems.