Abstract:
Due to the quantization noise created by analog-to-digital converters (ADCs) with just one bit of resolution, signal identification in Multiple-Input Multiple-Output (MIM...Show MoreMetadata
Abstract:
Due to the quantization noise created by analog-to-digital converters (ADCs) with just one bit of resolution, signal identification in Multiple-Input Multiple-Output (MIMO) systems is a very difficult task to perform. This is due to the fact that MIMO systems have many inputs in addition to a large number of outputs. The conventional methods of detection have a difficult time properly handling this noise. In this study, we describe a novel strategy based on deep learning for improving signal identification performance in MIMO systems like these that employ ADCs with just one bit of resolution. The method of deep learning that we recommend takes advantage of the power of neural networks to generate intricate connections from information that has been quantized to one bit. We provide the model with the ability to learn to differentiate between the several separate delivered signals while simultaneously decreasing the impacts of quantization noise by training it on a massive dataset consisting of MIMO channel realisations. This is done by providing it with the opportunity to learn from a variety of MIMO channel realisations. In-depth simulations are carried out in order to evaluate how well our deep learning-based approach to identifying signals works, and the results obtained from these simulations are compared to the findings obtained from more conventional detection methods. The results of our research indicate that deep learning algorithms perform better than traditional ones in terms of bit error rate (BER) and symbol recognition accuracy, particularly in challenging conditions with low signal-to-noise ratios (SNR).
Date of Conference: 18-20 September 2024
Date Added to IEEE Xplore: 15 January 2025
ISBN Information: