Navegando por Assunto "Compression"

Agora exibindo 1 - 4 de 4

Acesso aberto (Open Access)
Avaliação de técnicas de compressão de sinais Para o fronthaul
(Universidade Federal do Pará, 2019-11-27) BRITO, Flávio Mendes de; KLAUTAU JÚNIOR, Aldebaro Barreto da Rocha; http://lattes.cnpq.br/1596629769697284
The growing data demand of mobile networks has motivated the creation and evolution of architectures aiming to supply such transfer requirements. To meet these requirements, a number of challenges need to be met, including data transfer at the link between the Base Station Unit (BBU) and the Remote Radio Head (RRH). Known as fronthaul, this link requires high speed information transfer and one method that allows to tranfer more data using the same rate is data compression. Therefore, this study aimed to evaluate different techniques used in fronthaul data compression. Initially, the efficiency of some quantizers such as the scalar quantizer (SQ), twodimensional vector (VQ) and the Trellis Coded Quantization (TCQ) was verified. The analysis consists of combining these quantizers with resampling, Block Scaling and Huffman coding. In both analyzes, it was found that the system using TCQ as quantizer obtained the best relationship between Error Vector Magnitude (EVM) and computational cost, offering an EVM lower than the scalar quantizer and a computational cost lower than the vector quantizer.
Acesso aberto (Open Access)
Compressão de CSI para MIMO distribuído com processamento centralizado
(Universidade Federal do Pará, 2024-06-18) SILVA, Marcos Davi Lima da; RAMALHO, Leonardo Lira; http://lattes.cnpq.br/7565458988876048; https://orcid.org/0000-0003-3165-1941
In Distributed-MIMO (D-MIMO), a large number of distributed Antenna Points (APs) are coordinated by a Central Unit (CU) to serve a limited number of users with the same time/frequency resources, which brings improvements in spectral efficiency. The success of D-MIMO depends on precoding and power allocation, which can be performed completely centrally on the CU or distributed across APs. The centralized approach has greater spectral efficiency than the distributed implementation, but requires a significant spike in fronthaul traffic due to the exchange of Channel State Information (CSI) between APs and CU. In this work, CSI compression schemes are proposed to enable practical and centralized implementation of D-MIMO. It is shown that depending on the compression configuration, the spectral efficiency can be as good as in the case without compression. Furthermore, this work explores the implementation of multiple-input multiple-output (MIMO) within the framework of the New Radio (NR) architecture. The study evaluates a distributed MIMO deployment using NR signals with compression and evaluates its performance compared to the uncompressed scenario. Through simulations using the NR physical layer, the results also show that the spectral efficiency can be as good as in the uncompressed case depending on the compression configuration. Finally, the simulations with NR signals highlight important practical aspects and the feasibility of implementing D-MIMO in the 5G architecture and beyond 5G.
Acesso aberto (Open Access)
Compressão de sinais em sistemas de rádio sobre fibra digital para redes fronthaul
(Universidade Federal do Pará, 2019-07-23) MATE, Dércio Manuel; TEIXEIRA, António Luis de Jesus; OLIVEIRA, Rosinei de Sousa; http://lattes.cnpq.br/3853897074036715; COSTA, João Crisóstomo Weyl Albuquerque; http://lattes.cnpq.br/9622051867672434
The introduction of technologies such as Carrier Aggregation (CA), Massive Multiple Input Multiple Output (MIMO) and Coordinated Multipoint (CoMP), aiming to improve the performance of LTE and LTE-A systems, increases the challenge for deploying Mobile Fronthaul due to the network capacity limitation to support higher transmission rates. An approach to deal with Frontahul’s capacity limitation is data compression. Several techniques have been developed for signal compression in fronthaul, and most of these techniques compress the signal transmitted in baseband. In this work, a compression technique is developed for specific scenarios of Digital Radio-over-Fiber systems, transmitting the signal in intermediate frequency (IF). This technique uses the radio channel state information (CSI) to control signal compression in the fronthaul. The simulation results with the developed technique demonstrate its ability to reduce the data transmitted onthe network by 45.05%. In addition, this technique allows the transmission of 64 QAM modulated signals using a lower quantizer resolution, e.g., 4 bits per sample, maintaining the EVM below 3GPP recommended threshold (8%). Finally, the performance of the fronthaul network is evaluated experimentally in an optical link of 20-km, considering scenarios with and without signal compression.
Acesso aberto (Open Access)
Compression of activation signals from partitioned deep neural networks exploring temporal correlation
(Universidade Federal do Pará, 2024-11-27) SILVA, Lucas Damasceno; KLAUTAU JÚNIOR, Aldebaro Barreto da Rocha; http://lattes.cnpq.br/1596629769697284
The use of artificial neural networks for object detection, along with advancements in 6G and IoT research, plays an important role in applications such as drone-based monitoring of structures, search and rescue operations, and deployment on hardware platforms like FPGAs. However, a key challenge in implementing these networks on such hardware is the need to economize computational resources. Despite substantial advances in computational capacity, implementing devices with ample resources remains challenging. As a solution, techniques for partitioning and compressing neural networks, as well as compressing activation signals (or feature maps), have been developed. This work proposes a system that partitions neural network models for object detection in videos, allocating part of the network to an end device and the remainder to a cloud server. The system also compresses the feature maps generated by the last layers on the end device by exploiting temporal correlation, enabling a predictive compression scheme. This approach allows neural networks to be embedded in low-power devices while respecting the computational limits of the device, the transmission rate constraints of the communication channel between the device and server, and the network’s accuracy requirements. Experiments conducted on pre-trained neural network models show that the proposed system can significantly reduce the amount of data to be stored or transmitted by leveraging temporal correlation, facilitating the deployment of these networks on devices with limited computational power