Haoran Chang

Haoran Chang

Auto-generated profile, based on publicly available information

Haoran Chang

School of Artificial Intelligence, Beijing University of Posts and Telecommunications, Beijing, China

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
17 June 2022 in IEEE Transactions on Wireless Communications
Verified
Journal Article
Closed access
XLXin Liang
Xin Liang
HCHaoran Chang
Haoran Chang
HLHaozhen Li
Haozhen Li
XGXinyu Gu
Xinyu Gu
LZlin zhang
lin zhang
Deep learning (DL)-based channel state information (CSI) feedback improves the capacity and energy efficiency of massive multiple-input multiple-output (MIMO) systems in frequency division duplexing mode. However, multiple neural networks with different lengths of feedback overhead are required by time-varying bandwidth resources. The storage space required at the user equipment (UE) and the base station (BS) for these models increases linearly with the number of models. In this paper, we propose a DL-based changeable-rate framework with novel quantization scheme to improve the efficiency and feasibility of CSI feedback systems. This framework can reutilize all the network layers to achieve overhead-changeable CSI feedback to optimize the storage efficiency at the UE and the BS sides. Designed quantizer in this framework can avoid the normalization and gradient problems faced by traditional quantization schemes. Specifically, we propose two DL-based changeable-rate CSI feedback networks CH-CsiNetPro and CH-DualNetSph by introducing a feedback overhead control unit. Then, a pluggable quantization block (PQB) is developed to further improve the encoding efficiency of CSI feedback in an end-to-end way. Compared with existing CSI feedback methods, the proposed framework saves the storage space by about 50% with changeable-rate scheme and improves the encoding efficiency with the quantization module.

8 March 2022 in IEEE Wireless Communications Letters
Verified
Journal Article
Closed access
HLHaozhen Li
Haozhen Li
BZBoyuan Zhang
Boyuan Zhang
HCHaoran Chang
Haoran Chang
XLXin Liang
Xin Liang
XGXinyu Gu
Xinyu Gu
The deep learning-based (DL-based) channel state information (CSI) feedback in the frequency division duplexing (FDD) massive multiple-input multiple-output (MIMO) system has demonstrated its potential and efficiency. However, conventional neural networks cannot fully utilize the complex-valued nature of the downlink channel. In addition, multi-scale and multi-resolution features of CSI can be further explored. In this letter, we present a complex-valued lightweight neural network for CSI feedback named CVLNet. The CVLNet adopts the complex-valued neural network components in a multi-scale feature augmentation encoder and a multi-resolution X-shaped reconstruction decoder with a series of lightweight details. The experiment results show that the proposed CVLNet maintains the same-level parameters of the encoder with state-of-the-art (SOTA) lightweight networks while outperforming them with at most a 33.4% improvement in accuracy under severe compression rates.

8 October 2021 in IEEE Wireless Communications Letters
Verified
Journal Article
Closed access
HCHaoran Chang
Haoran Chang
XLXin Liang
Xin Liang
HLHaozhen Li
Haozhen Li
JSJinghan Shen
Jinghan Shen
XGXinyu Gu
Xinyu Gu
LZlin zhang
lin zhang
Deep learning (DL)-based channel state information (CSI) feedback algorithms for massive multiple-input multiple-output (MIMO) can provide high beamforming accuracy to improve the throughput. However, bitstream errors in the feedback process can significantly affect the performance of CSI reconstruction. In this letter, we focus on building high reconstruction accuracy CSI feedback algorithm in the presence of bitstream errors. Specifically, we firstly introduce a DL-based architecture named ATNet, which can improve at least 2.96dB Normalized Mean Square Error (NMSE) compared with the existing algorithm. Then, we propose an error correction block called ECBlock and a two-step training strategy. Compared with traditional methods, the proposed scheme can reduce the influence of quantization and bitstream errors more effectively and improve the reconstruction accuracy.