For millimeter wave (mmWave) massive multiple-input multiple-output (MIMO) systems, hybrid processing architecture is usually used to reduce the complexity and cost, which poses a very challenging issue in channel estimation. In this paper, deep convolutional neural network (CNN) is employed to address this problem. We first propose a spatial-frequency CNN (SF-CNN) based channel estimation exploiting both the spatial and frequency correlation, where the corrupted channel matrices at adjacent subcarriers are input into the CNN simultaneously. Then, exploiting the temporal correlation in time-varying channels, a spatial-frequency-temporal CNN (SFT-CNN) based approach is developed to further improve the accuracy. Moreover, we design a spatial pilot-reduced CNN (SPR-CNN) to save spatial pilot overhead for channel estimation, where channels in several successive coherence intervals are grouped and estimated by a channel estimation unit with memory. Numerical results show that the proposed SF-CNN and SFT-CNN based approaches outperform the non-ideal minimum mean-squared error (MMSE) estimator but with reduced complexity, and achieve the performance close to the ideal MMSE estimator that is very difficult to be implemented in practical situations. They are also robust to different propagation scenarios. The SPR-CNN based approach achieves comparable performance to SF-CNN and SFT-CNN based approaches while only requires about one third of spatial pilot overhead at the cost of complexity. Our work clearly shows that deep CNN can efficiently exploit channel correlation to improve the estimation performance for mmWave massive MIMO systems.
10 May 2014 in EURASIP Journal on Advances in Signal Processing
Generalized frequency division multiplexing (GFDM) is a block filtered multicarrier modulation scheme recently proposed for future wireless communication systems. It generalizes the concept of orthogonal frequency division multiplexing (OFDM), featuring multiple circularly pulse-shaped subsymbols per subcarrier. This paper presents an algorithm for GFDM synchronization and investigates the use of a preamble that consists of two identical parts combined with a windowing process in order to satisfy low out of band radiation requirements. The performance of time and frequency estimation, with and without windowing, is evaluated in terms of the statistical properties of residual offsets and the impact on symbol error rate over frequency-selective channels. A flexible metric that quantifies the penalty of misalignments is derived. The results show that this approach performs practically as state-of-the-art OFDM schemes known in the literature, while it additionally can reduce the sidelobes of the spectrum emission.
5GNOW is a European collaborative research project questioning the design targets of LTE and LTE-Advanced, in particular the obedience to strict synchronism and orthogonality. By introducing new waveforms, the project fosters a re-design of physical and partially MAC layer to support heterogeneous traffic (high rate, sporadic access, carrier aggregation). For system level performance evaluations of the non-orthogonal waveforms (together with key radio resource management algorithms) an abstract view of the physical layer with analytical models going beyond standard OFDM performance evaluation (e.g. due to additional dependencies between resource blocks) is needed. In this paper, elements of a system level simulation platform are developed, which provide necessary abstractions and sufficient accuracy to capture the essential physical layer behavior at a reasonable computational complexity.
Future wireless communication systems are demanding a more flexible physical layer. GFDM is a block filtered multicarrier modulation scheme proposed to add multiple degrees of freedom and cover other waveforms in a single framework. In this paper, GFDM modulation and demodulation will be presented as a frequency-domain circular convolution, allowing for a reduction of the implementation complexity when MF, ZF and MMSE filters are employed in the inner and outer receiver operation. Also, precoding is introduced to further increase GFDM flexibility, addressing a wider set of applications.
The next generation of wireless networks will face different challenges from new scenarios. The main contribution of this paper is to show that Generalized Frequency Division Multiplexing (GFDM), as a baseline of flexible circular filtered multicarrier systems, can be used as a framework to virtualize the PHY service for the upper layers of 5G networks. This framework opens the possibility to apply software-defined network principles to produce software-defined waveforms capable of addressing the requirements of future mobile networks. Hence, a block oriented concept will be used to provide the modulation service, emulating different flavors of waveforms designed to go beyond the well-established Orthogonal Frequency Division Multiplexing (OFDM) principles, in scenarios where they perform best. The virtual physical layer (PHY) service opens the opportunity to have a fast and dynamic evolution of the infrastructure, as applications change over time. The presented unified modulation concept contributes with future research directions to address burst and continuous transmissions, referencing basic approaches for synchronization and advanced receiver design that can be exploited in future for the whole frame structure design and channel estimation strategies.
For millimeter wave (mmWave) massive multiple-input multiple-output (MIMO) systems, hybrid processing architecture is usually used to reduce the complexity and cost, which poses a very challenging issue in channel estimation. In this paper, deep convolutional neural network (CNN) is employed to address this problem. We first propose a spatial-frequency CNN (SF-CNN) based channel estimation exploiting both the spatial and frequency correlation, where the corrupted channel matrices at adjacent subcarriers are input into the CNN simultaneously. Then, exploiting the temporal correlation in time-varying channels, a spatial-frequency-temporal CNN (SFT-CNN) based approach is developed to further improve the accuracy. Moreover, we design a spatial pilot-reduced CNN (SPR-CNN) to save spatial pilot overhead for channel estimation, where channels in several successive coherence intervals are grouped and estimated by a channel estimation unit with memory. Numerical results show that the proposed SF-CNN and SFT-CNN based approaches outperform the non-ideal minimum mean-squared error (MMSE) estimator but with reduced complexity, and achieve the performance close to the ideal MMSE estimator that is very difficult to be implemented in practical situations. They are also robust to different propagation scenarios. The SPR-CNN based approach achieves comparable performance to SF-CNN and SFT-CNN based approaches while only requires about one third of spatial pilot overhead at the cost of complexity. Our work clearly shows that deep CNN can efficiently exploit channel correlation to improve the estimation performance for mmWave massive MIMO systems.