Lijun Ge
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Lijun Ge has a total of 34 co-authors. Here are the top co-authors by number of shared publications:
- Gaojie Chen
Sun Yat-sen University
4 shared publications - Lizheng Wang
Tianjin Key Laboratory of Optoelectronic Detection Technology and System and the School of Electronics and Information Engineering, Tiangong University, Tianjin, China
4 shared publications Bo Dai
Tianjin Key Laboratory of Optoelectronic Detection Technology and System, School of Electrical and Electronic Engineering, Tiangong University, Tianjin, China
2 shared publications- Feifei Gao
Tsinghua University
2 shared publications - Jun Tong
University of Wollongong
2 shared publications Shixun Niu
School of Electronics and Information Engineering, and the Tianjin Key Laboratory of Optoelectronic Detection Technology and System, Tiangong University, Tianjin, China
2 shared publications
Latest Publications
- 13 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Green Communications and Networking
- Vol. 10, 2206-2219
Channel Estimation for mmWave mMIMO Systems With Mixed-ADC: Conditional Generative Adversarial Network and Deep Transfer Learning
Millimeter wave (mmWave) massive multiple input multiple output (mMIMO) systems have the advantages of abundant spectrum resources, high transmission rate, and high multiplexing gain. However, the large number of antennas leads to high power consumption and difficult channel estimation. Although the problem of high power consumption can be addressed by reducing the analog-to-digital converter (ADC) resolution of radio frequency (RF) links, considering the accuracy of channel state information (CSI), high-resolution antennas are still needed to be deployed on a small number of RF terminals. Therefore, a mixed-resolution ADC should be used for signal transmission to consider channel estimation and power consumption tradeoff. To reduce the influence of quantization noise generated by the low- resolution ADC on channel estimation performance, we propose two deep learning-based channel estimation methods for mmWave mMIMO systems with mixed-resolution ADCs. Based on the conditional generative adversarial network, the first method has the advantage of generating more useful CSI with less available high-resolution CSI, which fits the characteristics of the mixed ADC system. Moreover, the second proposed method exploits deep transfer learning to train low-resolution antenna data using network parameters that have been trained at high-resolution antennas. Both methods can fully utilize the information from high-resolution antennas while mining the effective information from low-resolution antennas, thereby achieving high-resolution channel estimation while reducing mMIMO system overhead. Simulation results demonstrate that both methods can perform better than deep neural networks in mixed-resolution ADC scenarios.
- 13 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Green Communications and Networking
- Vol. 10, 2206-2219
Channel Estimation for mmWave mMIMO Systems With Mixed-ADC: Conditional Generative Adversarial Network and Deep Transfer Learning
Millimeter wave (mmWave) massive multiple input multiple output (mMIMO) systems have the advantages of abundant spectrum resources, high transmission rate, and high multiplexing gain. However, the large number of antennas leads to high power consumption and difficult channel estimation. Although the problem of high power consumption can be addressed by reducing the analog-to-digital converter (ADC) resolution of radio frequency (RF) links, considering the accuracy of channel state information (CSI), high-resolution antennas are still needed to be deployed on a small number of RF terminals. Therefore, a mixed-resolution ADC should be used for signal transmission to consider channel estimation and power consumption tradeoff. To reduce the influence of quantization noise generated by the low- resolution ADC on channel estimation performance, we propose two deep learning-based channel estimation methods for mmWave mMIMO systems with mixed-resolution ADCs. Based on the conditional generative adversarial network, the first method has the advantage of generating more useful CSI with less available high-resolution CSI, which fits the characteristics of the mixed ADC system. Moreover, the second proposed method exploits deep transfer learning to train low-resolution antenna data using network parameters that have been trained at high-resolution antennas. Both methods can fully utilize the information from high-resolution antennas while mining the effective information from low-resolution antennas, thereby achieving high-resolution channel estimation while reducing mMIMO system overhead. Simulation results demonstrate that both methods can perform better than deep neural networks in mixed-resolution ADC scenarios.
- 11 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 75 (6) , 10542-10553
Enhancing Channel Estimation in Millimeter-Wave Massive MIMO Systems via Nested Planar Arrays With Joint 2-D DOA Estimation
In millimeter-wave massive multiple-input multiple-output (MIMO) systems, the limited number of radio frequency chains and large antenna arrays at the receiver result in challenges for channel estimation, including increased cost and system. These limitations consequently impair the accuracy and effectiveness of channel estimation in these systems. To mitigate these problems, we consider the deployment of two-dimensional nested planar arrays (NPA) at base stations. In traditional MIMO systems, phase shifters are typically employed to adjust the phase of individual antenna elements for beamforming and antenna selection. In this design, we introduce a switching network instead of a phase shifter. The switching network achieves antenna selection by dynamically switching between different antenna combinations. This approach simplifies hardware design and reduces the power consumption associated with phase shifters. The proposed methodology parameterizes each channel path through its direction of arrival (DOA) and channel gain. By leveraging the virtual array elements within NPA, we propose a joint spatial smoothing multiple signal classification algorithm for DOA estimation, which can effectively remove the phase interference caused by multipath by performing multiple smoothing processes in different directions, thus improving the angle estimation capability. Subsequently, the channel gain of each path is estimated using the least squares method. The simulation outcomes validated the effectiveness of the proposed NPA configuration and its corresponding channel estimation method. In addition, the proposed method is superior in terms of both resolution and accuracy of channel estimation compared to the traditional methods.
- 11 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 75 (6) , 10542-10553
Enhancing Channel Estimation in Millimeter-Wave Massive MIMO Systems via Nested Planar Arrays With Joint 2-D DOA Estimation
In millimeter-wave massive multiple-input multiple-output (MIMO) systems, the limited number of radio frequency chains and large antenna arrays at the receiver result in challenges for channel estimation, including increased cost and system. These limitations consequently impair the accuracy and effectiveness of channel estimation in these systems. To mitigate these problems, we consider the deployment of two-dimensional nested planar arrays (NPA) at base stations. In traditional MIMO systems, phase shifters are typically employed to adjust the phase of individual antenna elements for beamforming and antenna selection. In this design, we introduce a switching network instead of a phase shifter. The switching network achieves antenna selection by dynamically switching between different antenna combinations. This approach simplifies hardware design and reduces the power consumption associated with phase shifters. The proposed methodology parameterizes each channel path through its direction of arrival (DOA) and channel gain. By leveraging the virtual array elements within NPA, we propose a joint spatial smoothing multiple signal classification algorithm for DOA estimation, which can effectively remove the phase interference caused by multipath by performing multiple smoothing processes in different directions, thus improving the angle estimation capability. Subsequently, the channel gain of each path is estimated using the least squares method. The simulation outcomes validated the effectiveness of the proposed NPA configuration and its corresponding channel estimation method. In addition, the proposed method is superior in terms of both resolution and accuracy of channel estimation compared to the traditional methods.
- 18 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Green Communications and Networking
- Vol. 10, 859-868
Channel Estimation for mmWave mMIMO Systems With Low-Resolution ADCs Based on Adaptive Quantization Improved CGAN
Millimeter wave massive multiple-input multiple-output systems offer higher information capacity and can mitigate severe propagation losses. However, the rapid increase in the number of antennas results in higher hardware costs and power consumption. Although these issues can be addressed by reducing the resolution of analog-to-digital converters (ADCs) in the radio frequency chain, low-resolution ADCs introduce significant nonlinear errors when quantizing received signals, making accurate channel estimation very challenging. To mitigate the impact of quantization noise from low-resolution ADCs on channel estimation performance, we propose a novel channel estimation method based on adaptive quantization conditional generative adversarial network (AQ-CGAN). Instead of using traditional quantization methods, we convert the quantized received signals and pilot sequences as well as the quantization thresholds into two-dimensional images, randomly generate and iteratively optimize the thresholds under the CGAN framework. At the same time, we use the neural network and CGAN to establish a mapping relationship between quantization thresholds and channel matrices to achieve adaptive quantization, and introduce a new loss function to improve the channel estimation accuracy in massive MIMO systems. Simulation results demonstrate that the proposed algorithm outperforms traditional quantized CGAN channel estimation under the same signal-to-noise ratio conditions.
- 18 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Green Communications and Networking
- Vol. 10, 859-868
Channel Estimation for mmWave mMIMO Systems With Low-Resolution ADCs Based on Adaptive Quantization Improved CGAN
Millimeter wave massive multiple-input multiple-output systems offer higher information capacity and can mitigate severe propagation losses. However, the rapid increase in the number of antennas results in higher hardware costs and power consumption. Although these issues can be addressed by reducing the resolution of analog-to-digital converters (ADCs) in the radio frequency chain, low-resolution ADCs introduce significant nonlinear errors when quantizing received signals, making accurate channel estimation very challenging. To mitigate the impact of quantization noise from low-resolution ADCs on channel estimation performance, we propose a novel channel estimation method based on adaptive quantization conditional generative adversarial network (AQ-CGAN). Instead of using traditional quantization methods, we convert the quantized received signals and pilot sequences as well as the quantization thresholds into two-dimensional images, randomly generate and iteratively optimize the thresholds under the CGAN framework. At the same time, we use the neural network and CGAN to establish a mapping relationship between quantization thresholds and channel matrices to achieve adaptive quantization, and introduce a new loss function to improve the channel estimation accuracy in massive MIMO systems. Simulation results demonstrate that the proposed algorithm outperforms traditional quantized CGAN channel estimation under the same signal-to-noise ratio conditions.
- 8 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Cognitive Communications and Networking
- Vol. 12, 2410-2423
Deep Learning-Based Channel Estimation for Massive MIMO: A Sparsity Adaptive Compressive Sensing Method and FPGA Implementation
Massive multiple-input multiple-output (MIMO) systems face significant overhead due to the surge in the number of antennas, posing notable challenges for hardware implementation. Currently, channel estimation methods that combine Compressed Sensing (CS) with Deep Learning (DL) are developed to address this issue. However, the sparsity adaptation of channel estimation methods is still a huge problem, and Channel State Information (CSI) estimation accuracy still needs improvement. This paper proposes a novel structure combining CS and DL, named the sparsity adaptive deep learning compressed sensing (Sa-DLCS) method, which can leverage sparsity adaptation to obtain globally optimal channel estimation. It includes three key steps: 1) preprocessing, 2) updating the neural network (NN) mapping relation, and 3) postprocessing. First, sparsity adaptation is realized to select matching atoms by taking the arithmetic mean value and the harmonic mean value of the correlation between the received signal and the measurement matrix as the threshold. Second, it updates the NN input-output mapping relationship. Consequently, the NN has a new mapping relationship involving the measurement matrix in the CS. Third, based on the output of the NN, a high-precision channel estimation is employed, where the errors are related to the error size of the inverse of the measurement matrix. Simulation results showed that the proposed methods have better normalized mean square error performance than the existing deep learning based compressed sensing method. This paper also provides the hardware architecture of Sa-DLCS algorithm and discusses the optimization strategies for its hardware implementation on Field-Programmable Gate Array, indicating Sa-DLCS is a hardware-efficient channel estimation algorithm. In addition, we implement Sa-DLCS algorithm on xilinx Virtex-7 FPGA, and compare it with the hardware implementation scheme of advanced compressed sensing based channel estimator. The results show that the proposed Sa-DLCS channel estimation algorithm can reduce the reconstruction time by up to 54% without increasing the FPGA logic resources.
- 8 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Cognitive Communications and Networking
- Vol. 12, 2410-2423
Deep Learning-Based Channel Estimation for Massive MIMO: A Sparsity Adaptive Compressive Sensing Method and FPGA Implementation
Massive multiple-input multiple-output (MIMO) systems face significant overhead due to the surge in the number of antennas, posing notable challenges for hardware implementation. Currently, channel estimation methods that combine Compressed Sensing (CS) with Deep Learning (DL) are developed to address this issue. However, the sparsity adaptation of channel estimation methods is still a huge problem, and Channel State Information (CSI) estimation accuracy still needs improvement. This paper proposes a novel structure combining CS and DL, named the sparsity adaptive deep learning compressed sensing (Sa-DLCS) method, which can leverage sparsity adaptation to obtain globally optimal channel estimation. It includes three key steps: 1) preprocessing, 2) updating the neural network (NN) mapping relation, and 3) postprocessing. First, sparsity adaptation is realized to select matching atoms by taking the arithmetic mean value and the harmonic mean value of the correlation between the received signal and the measurement matrix as the threshold. Second, it updates the NN input-output mapping relationship. Consequently, the NN has a new mapping relationship involving the measurement matrix in the CS. Third, based on the output of the NN, a high-precision channel estimation is employed, where the errors are related to the error size of the inverse of the measurement matrix. Simulation results showed that the proposed methods have better normalized mean square error performance than the existing deep learning based compressed sensing method. This paper also provides the hardware architecture of Sa-DLCS algorithm and discusses the optimization strategies for its hardware implementation on Field-Programmable Gate Array, indicating Sa-DLCS is a hardware-efficient channel estimation algorithm. In addition, we implement Sa-DLCS algorithm on xilinx Virtex-7 FPGA, and compare it with the hardware implementation scheme of advanced compressed sensing based channel estimator. The results show that the proposed Sa-DLCS channel estimation algorithm can reduce the reconstruction time by up to 54% without increasing the FPGA logic resources.
- Open Access
- 19 October 2023
- Published by Institution of Engineering and Technology (IET) in IET Communications
- Vol. 17 (19) , 2152-2161
Mixed RNN‐DNN based channel prediction for massive MIMO‐OFDM systems
Channel state information (CSI), which is crucial for resource allocation and system performance in time division duplex (TDD) massive multiple‐input multiple‐output (MIMO) systems, is difficult to predict because of the channel's time‐varying nature. To overcome this limitation, a scheme for channel prediction combined with deep learning (DL) is proposed. The system uses a deep neural network (DNN) to interpolate channel estimates from a few received pilot signals and a recurrent neural network (RNN) to train through the current time and the recent historical channel estimates to predict the CSI while the channel is constantly varying. In the end, a mixed neural network of RNN and DNN, is called MRDNN. In addition, the proposed DL‐based method does not rely on the relevant feature information about the channel, such as internal characteristics and parameters of the channel itself or channel statistical information, which improves its effectiveness in practical applications. The results of the simulation show that the MRDNN‐based method is better than the existing methods, like traditional AR method and NL Kalman method, and also can be effective in improving the quality of channel prediction and the performance of the system under the dynamic change scenario of low mobility.
- 19 October 2023
- Published by Institution of Engineering and Technology (IET) in IET Communications
- Vol. 17 (19) , 2152-2161
Open AccessMixed RNN‐DNN based channel prediction for massive MIMO‐OFDM systems
Channel state information (CSI), which is crucial for resource allocation and system performance in time division duplex (TDD) massive multiple‐input multiple‐output (MIMO) systems, is difficult to predict because of the channel's time‐varying nature. To overcome this limitation, a scheme for channel prediction combined with deep learning (DL) is proposed. The system uses a deep neural network (DNN) to interpolate channel estimates from a few received pilot signals and a recurrent neural network (RNN) to train through the current time and the recent historical channel estimates to predict the CSI while the channel is constantly varying. In the end, a mixed neural network of RNN and DNN, is called MRDNN. In addition, the proposed DL‐based method does not rely on the relevant feature information about the channel, such as internal characteristics and parameters of the channel itself or channel statistical information, which improves its effectiveness in practical applications. The results of the simulation show that the MRDNN‐based method is better than the existing methods, like traditional AR method and NL Kalman method, and also can be effective in improving the quality of channel prediction and the performance of the system under the dynamic change scenario of low mobility.
- 11 October 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Systems Journal
- Vol. 16 (3) , 4675-4685
Deep Neural Network Based Channel Estimation for Massive MIMO-OFDM Systems With Imperfect Channel State Information
Massive multi-input multi-output (MIMO) has aroused extensive interest in communication industry, as it can effectively increase communication system capacity and reduce transmit power. For massive MIMO-orthogonal frequency division multiplexing (OFDM) systems, traditional pilot assisted channel estimation methods need to obtain the channel response of data subcarriers by interpolation. However, when the channel response of continuous data subcarriers changes dramatically, the accuracy of channel estimation will be decreased. In this article, therefore, a novel channel prediction framework that integrates the imperfect channel estimation of the massive MIMO-OFDM into the deep neural network scheme is proposed, thus replacing the interpolation method and achieving higher accuracy of channel estimation. From algorithm analysis and simulation results, the performance of proposed deep neural network channel prediction method based on the imperfect channel estimation outperforms the conventional least square method based on interpolation. Furthermore, compared with minimum mean square error, there are no need for channel autocorrelation matrix, prior statistics of noise variance, and the complex matrix inversion operations. In summary, deep neural network is an efficient method for the imperfect channel estimation in massive MIMO-OFDM systems.
- 11 October 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Systems Journal
- Vol. 16 (3) , 4675-4685
Deep Neural Network Based Channel Estimation for Massive MIMO-OFDM Systems With Imperfect Channel State Information
Massive multi-input multi-output (MIMO) has aroused extensive interest in communication industry, as it can effectively increase communication system capacity and reduce transmit power. For massive MIMO-orthogonal frequency division multiplexing (OFDM) systems, traditional pilot assisted channel estimation methods need to obtain the channel response of data subcarriers by interpolation. However, when the channel response of continuous data subcarriers changes dramatically, the accuracy of channel estimation will be decreased. In this article, therefore, a novel channel prediction framework that integrates the imperfect channel estimation of the massive MIMO-OFDM into the deep neural network scheme is proposed, thus replacing the interpolation method and achieving higher accuracy of channel estimation. From algorithm analysis and simulation results, the performance of proposed deep neural network channel prediction method based on the imperfect channel estimation outperforms the conventional least square method based on interpolation. Furthermore, compared with minimum mean square error, there are no need for channel autocorrelation matrix, prior statistics of noise variance, and the complex matrix inversion operations. In summary, deep neural network is an efficient method for the imperfect channel estimation in massive MIMO-OFDM systems.
- 16 November 2020
- Published by Taylor & Francis in Journal of the Chinese Institute of Engineers
- Vol. 43 (8) , 851-855
An improved MASH sigma-delta modulator with all-anolog interstage and input-feedforward architecture
This paper puts forward an improved multi-stage noise-shaping (MASH) sigma-delta modulator. Derived from a fourth-order two-stage MASH architecture with adder-less architecture of cascade of integrators with feedforward (CIFF), the proposed modulator improves the method of extracting quantization noise of the first stage to form an all-analog interstage, and two input-feedforward paths are added to the second stage to reduce the output swing of the integrator. Analyses and measurements indicate that the proposed architecture has lower in-band quantization noise power, higher signal-to-quantization noise ratio (SQNR), and lower power consumption and complexity as well. When operating from a 1.2 V supply, the modulator achieves 107.56 dB peak SQNR, 71.8 dB peak Signalnoise and distortion ratio while consuming 26 mW, with an OSR of 8 at a 160 MHz sampling frequency.
- 16 November 2020
- Published by Taylor & Francis in Journal of the Chinese Institute of Engineers
- Vol. 43 (8) , 851-855
An improved MASH sigma-delta modulator with all-anolog interstage and input-feedforward architecture
This paper puts forward an improved multi-stage noise-shaping (MASH) sigma-delta modulator. Derived from a fourth-order two-stage MASH architecture with adder-less architecture of cascade of integrators with feedforward (CIFF), the proposed modulator improves the method of extracting quantization noise of the first stage to form an all-analog interstage, and two input-feedforward paths are added to the second stage to reduce the output swing of the integrator. Analyses and measurements indicate that the proposed architecture has lower in-band quantization noise power, higher signal-to-quantization noise ratio (SQNR), and lower power consumption and complexity as well. When operating from a 1.2 V supply, the modulator achieves 107.56 dB peak SQNR, 71.8 dB peak Signalnoise and distortion ratio while consuming 26 mW, with an OSR of 8 at a 160 MHz sampling frequency.
- 1 August 2016
- Published by Elsevier in Free Radical Biology and Medicine
- Vol. 97, 179-191
Peroxiredoxin-2 nitrosylation facilitates cardiomyogenesis of mouse embryonic stem cells via XBP-1s/PI3K pathway
- 1 August 2016
- Published by Elsevier in Free Radical Biology and Medicine
- Vol. 97, 179-191
Peroxiredoxin-2 nitrosylation facilitates cardiomyogenesis of mouse embryonic stem cells via XBP-1s/PI3K pathway