Complexity Channel Estimation Research Articles

Characteristics of Channel Spreading Function and Performance of OTFS in High-Speed Railway

Orthogonal time frequency space (OTFS) modulation is an emerging technology to tackle time-frequency (TF) selective channel in high mobility scenarios. In OTFS, resource is multiplexed in the delay-Doppler (DD) domain. Based on the potential sparsity, separability, stability and compactness of the channel spreading function, OTFS is able to realize lower complexity of channel estimation, higher diversity and higher reliability compared with orthogonal frequency division multiplexing (OFDM). However, the channel spreading function for practical communication systems is rarely considered in the current OTFS-related literature. High-speed railway (HSR) is a typical high mobility scenario with trains travelling at over 200km/h, which has the potential to employ OTFS. To this end, the HSR channel spreading function is characterized and the performance of OTFS in HSR is evaluated based on the realistic channel measurement in this article. Firstly, the HSR channel in TF domain is measured based on the long term evolution railway (LTE-R) network. Then, the characteristics of the channel spreading function are analyzed. In particular, the impact of time domain channel fading on the spreading function is investigated. The characteristics of the measured spreading function are analyzed with the proposed metrics in railway viaduct and tunnel scenarios. Based on the above analysis, an algorithm for generating the channel spreading function is proposed. Feasibility of the proposed DD domain channel generation algorithm is verified through comparing metrics of which to those of the measured channel. By simulating the bit error rate (BER) and mean square channel estimation error performances of OTFS modulation in the practical band-limited systems, it is shown that the impacts of Doppler shift, delay, SFFT and time domain channel fading need be considered for the application of OTFS modulation, in contrast to the state-of-art DD domain channel generation scheme based on tap delay link (TDL) model. For example, compared to OTFS modulation under the ideal channel spreading function, OTFS modulation requires a signal gain greater than 5 dB under the practical channel spreading function affected by above factors, to achieve the same BER less than 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-2</sup> under the parameters and algorithms given in section V.

Cluster-aware channel estimation with deep learning method in deep-water acoustic communications.

In underwater acoustic (UWA) communications, channels often exhibit a clustered-sparse structure, wherein most of the channel impulse responses are near zero, and only a small number of nonzero taps assemble to form clusters. Several algorithms have used the time-domain sparse characteristic of UWA channels to reduce the complexity of channel estimation and improve the accuracy. Employing the clustered structure to enhance channel estimation performance provides another promising research direction. In this work, a deep learning-based channel estimation method for UWA orthogonal frequency division multiplexing (OFDM) systems is proposed that leverages the clustered structure information. First, a cluster detection model based on convolutional neural networks is introduced to detect the cluster of UWA channels. This method outperforms the traditional Page test algorithm with better accuracy and robustness, particularly in low signal-to-noise ratio conditions. Based on the cluster detection model, a cluster-aware distributed compressed sensing channel estimation method is proposed, which reduces the noise-induced errors by exploiting the joint sparsity between adjacent OFDM symbols and limiting the search space of channel delay spread. Numerical simulation and sea trial results are provided to illustrate the superior performance of the proposed approach in comparison with existing sparse UWA channel estimation methods.

Deep Learning-Based Soft Iterative-Detection of Channel-Coded Compressed Sensing-Aided Multi-Dimensional Index Modulation

The concept of Index Modulation (IM) has been actively researched as a benefit of its flexible trade-off between performance, achievable rate, energy efficiency, hardware cost and complexity. In order to fully exploit its the degrees of freedom, the concept of Multi-dimensional IM (MIM) has been developed in literature, where Compressed Sensing (CS) is often utilized to exploit the sparsity of the multi-dimensional transmitted signals. However, this flexibility and performance gains are attained at the cost of a substantially increased detection complexity. In this paper, we propose Deep Learning (DL) based detection for CS-aided MIM (CS-MIM), where both Hard-Decision (HD) as well as Soft-Decision (SD) detection combined with iterative decoding are conceived. More explicitly, firstly, we propose learning aided hard and soft detection for CS-MIM. Secondly, two novel neural network aided methods are proposed for Iterative Soft Detection (ISD), where iterations are carried out between the CS-MIM detector and a channel decoder. In contrast to the conventional detection of CS-MIM system, which critically relies on the knowledge of Channel State Information (CSI) at the receiver, the proposed learning-aided methods are capable of eliminating the overhead and complexity of Channel Estimation (CE), which results in an improved transmission rate. Explicitly, we develop an advanced DL architecture for blind-detection-aided MIM for the first time in the open literature, where the HD and SD CS algorithms are implemented by learning without the need for CE. Our simulation results demonstrate that the proposed learning techniques conceived for SD CS-MIM combined with iterative detection are capable of achieving near-capacity performance at a reduced complexity compared to the conventional model-based SD relying on CSI acquisition.

A multi-fusion integrated end-to-end deep kernel CNN based channel estimation for hybrid range UM-MIMO 6G communication systems

In today’s world, Terahertz Communication (THzCom) is envisioned as a highly promising technology to characterize 6G wireless communication within the range from 0.1 to 10 THz. This will configure emerging wireless applications due to the demand for their unique potential to support terabit-per-second transmission and effectively support ultra-high quality of service requirements. Aiming at massive dual beamforming problem in THzCom, the Ultra-Massive-Multiple Input and Multiple Output (UM-MIMO) with less array based hybrid channel modelling is adopted as the promising solutions. Further, in the case of an ultra-massive number of antennas, the beamforming matrix element is quite high-dimensional. This enlarged dimension of channel contributes to the high complexity of Channel Estimation (CE). To focus on these challenges, we presented a Hybrid Short and Long range Channel Model (HSLCM) based Multi-Fusion based Deep Kernel Convolutional Neural Network (MF-DKCNN) for THz UM-MIMO CE method. HSLCM channel modelling is used to determine the channel parameters of THz propagation paths, such as path loss, azimuth and elevation angles, distance, and the steering vector of the lens array. The MF integrated with the DKCNN model is used to estimate the noise-free received signal from accurate channel parameters and handles high-dimensional matrices with low complexity. MF employs the Channel Intersection Fusion Rule (CI-FR) to combine the received signal and channel matrix for a consistent fusion estimate of the known channel parameters from the system model. The DKCNN network structure has fifteen layers in which the MF estimates the noise-free channel matrices to be processed as training labels. Then, the kernel evaluation is used between layers to solve high dimensions and feature complexity. The proposed model is analysed in an outdoor THz channel scenario, which was simulated on the MATLAB (R2018A) platform. Hence, the evaluation results demonstrate that the proposed method outperforms high CE results with low complexity than other methods in terms of training data and Normalized Mean Square Error (NMSE).

Channel Tracking for RIS-Aided mmWave Communications Under High Mobility Scenarios

The emerging reconfigurable intelligent surface (RIS) technology is promising for applications in the millimeter wave (mmWave) communication systems to effectively compensate for propagation loss or tackle the blockage issue. Considering the high mobility of users in realistic scenarios, it is essential to adjust the phase shifts in real time to align the beam towards the mobile users, which requires to frequently estimate the channel state information. Hence, it is imperative to design efficient channel tracking schemes to avoid the complex channel estimation procedure. In this paper, we develop a novel channel tracking scheme with two advantages over conventional schemes. First, our tracking scheme is based on the cascaded angles at the RIS instead of the accurate angle values, which is more practical. Second, it can be employed under a more general setting where the noise can be non-Gaussian. Simulation results show the high tracking accuracy of our proposed scheme, and validate the superiority to the existing EKF-based tracking scheme.

Deep Learning for Parametric Channel Estimation in Massive MIMO Systems

Massive Multiple-Input Multiple-Output (MIMO) communication with a low bit error rate depends upon the availability of accurate Channel State Information (CSI) at the base station. The massive MIMO systems can be either deployed using time division duplexing with channel reciprocity assumption or by availing frequency division duplexing, which requires closed-loop feedback for CSI acquisition. The channel reciprocity simplifies transmission in time division duplexing; however, it suffers a bottleneck due to pilot contamination, whereas transmission in frequency division duplexing is challenged by channel estimation complexity, CSI feedback, and overall delay in CSI transfer. This paper proposes a simplified parametric channel model, its deep neural network aided estimation along with pilot decontamination for time division duplexing and a low rate parametric feedback and improved precoding for frequency division duplexing based massive MIMO systems. This novel framework integrates the massive MIMO parametric estimation and deep learning for improved estimation and precoding. Our proposed model also offers a unified approach for CSI acquisition with a performance bound on channel correlation in fast time-varying conditions. A theoretical model has been presented using Gaussian assumptions and validated by Monte-Carlo simulations. The results show total nullification of pilot contamination and high-performance gains when the proposed technique is employed for estimation.

Design of a Reconfigurable Intelligent Surface- Assisted FM-DCSK-SWIPT Scheme With Non-Linear Energy Harvesting Model

In this paper, we propose a reconfigurable intelligent surface (RIS)-assisted frequency-modulated (FM) differential chaos shift keying (DCSK) scheme with simultaneous wireless information and power transfer (SWIPT), called RIS-FM-DCSK-SWIPT scheme, for low-power, low-cost, and high-reliability wireless communication networks. In particular, the proposed scheme is developed under a non-linear energy-harvesting (EH) model which can accurately characterize the practical situation. The proposed RIS-FM-DCSK-SWIPT scheme has an appealing feature that it does not require channel state information, thus avoiding the complex channel estimation. We further derive the closed-form theoretical expressions for the energy shortage probability and bit error rate (BER) of the proposed scheme over the multipath Rayleigh fading channel. In addition, we investigate the influence of key parameters on the performance of the proposed transmission scheme in two different scenarios, i.e., RIS-assisted access point (RIS-AP) and dual-hop communication (RIS-DH). Finally, we carry out various Monte-Carlo experiments to verify the accuracy of the theoretical derivation, illustrate the performance advantage of the proposed scheme, and give some design insights for future study.

Deep Learning Multi-Class Approach for Human Fall Detection Based on Doppler Signatures.

Falling events are a global health concern with short- and long-term physical and psychological implications, especially for the elderly population. This work aims to monitor human activity in an indoor environment and recognize falling events without requiring users to carry a device or sensor on their bodies. A sensing platform based on the transmission of a continuous wave (CW) radio-frequency (RF) probe signal was developed using general-purpose equipment. The CW probe signal is similar to the pilot subcarriers transmitted by commercial off-the-shelf WiFi devices. As a result, our methodology can easily be integrated into a joint radio sensing and communication scheme. The sensing process is carried out by analyzing the changes in phase, amplitude, and frequency that the probe signal suffers when it is reflected or scattered by static and moving bodies. These features are commonly extracted from the channel state information (CSI) of WiFi signals. However, CSI relies on complex data acquisition and channel estimation processes. Doppler radars have also been used to monitor human activity. While effective, a radar-based fall detection system requires dedicated hardware. In this paper, we follow an alternative method to characterize falling events on the basis of the Doppler signatures imprinted on the CW probe signal by a falling person. A multi-class deep learning framework for classification was conceived to differentiate falling events from other activities that can be performed in indoor environments. Two neural network models were implemented. The first is based on a long-short-term memory network (LSTM) and the second on a convolutional neural network (CNN). A series of experiments comprising 11 subjects were conducted to collect empirical data and test the system's performance. Falls were detected with an accuracy of 92.1% for the LSTM case, while for the CNN, an accuracy rate of 92.1% was obtained. The results demonstrate the viability of human fall detection based on a radio sensing system such as the one described in this paper.

Multi-scenario millimeter wave wireless channel measurements and sparsity analysis

Wireless channel characteristics have significant impacts on channel modeling, estimation, and communication performance. While the channel sparsity is an important characteristic of wireless channels. Utilizing the sparse nature of wireless channels can reduce the complexity of channel modeling and estimation, and improve system design and performance analysis. Compared with the traditional sub-6 GHz channel, millimeter wave (mmWave) channel has been considered to be more sparse in existing researches. However, most research only assume that the mmWave channel is sparse, without providing quantitative analysis and evaluation. Therefore, this paper evaluates the sparsity of mmWave channels based on mmWave channel measurements. A vector network analyzer (VNA)-based mmWave channel sounder is developed to measure the channel at 28 GHz, and multi-scenario channel measurements are conducted. The Gini index, Rician K factor and root-mean-square (RMS) delay spread are used to measure channel sparsity. Then, the key factors affecting mmWave channel sparsity are explored. It is found that antenna steering direction and scattering environment will affect the sparsity of mmWave channel. In addition, the impact of channel sparsity on channel eigenvalue and capacity is evaluated and analyzed.

Belief-Propagation-Based Low-Complexity Channel Estimation and Detection for Underwater Acoustic Communications With Moving Transceivers

Achieving reliable communications with low complexity is challenging for underwater acoustic communications with moving transceivers, where the time-varying channels need to be estimated and tracked accurately and data detection needs to be performed with low complexity. In this article, with the use of a superimposed training (ST) scheme, we address this challenge by developing a low-complexity channel estimation and tracking algorithm, which is then integrated with low-complexity data detection in the frequency domain. ST is used to acquire improved channel-tracking capability. Based on belief propagation, we design a message-passing-based low-complexity bidirectional channel estimation (LCE-MP) algorithm, where all computational intensive parts are handled by the fast Fourier transform (FFT) algorithm, thereby achieving very efficient implementation with logarithmic complexity. Specifically, a message-passing-based fast information collection algorithm is presented to acquire “local” channel estimates, followed by the fusion of local channel estimates to achieve a “global” estimate of the channel. It is shown that the computational complexity per channel tap is only in a logarithmic level for the channel estimation and tracking. Moreover, the ST-scheme-based LCE-MP algorithm is combined with FFT-based data detection and decoding, which are performed in an iterative manner. The overall complexity of channel estimation and data detection is in a logarithmic level, and the system delivers excellent performance thanks to the joint processing. Field experiments were carried out in Jiaozhou Bay in 2019, and the experimental results verify the effectiveness of the proposed technique.

Survey on positioning information assisted mmWave beamforming training

mmWave communication is considered a major player for 5 G and beyond networks due to its wide unlicensed bandwidth. However, it comes with some challenges such as propagation and penetration losses. To overcome these issues, antenna beamforming is adopted. But it is difficult to estimate massive mmWave multiple input multiple output channel for active beamforming, so training using switched antenna arrays with structured codebook is used to perform mmWave beamforming training (BT). The conventional exhaustive search BT scheme wastes a long time and power, especially in pencil beams scenarios, hence, for a better communication, these issues have to be handled. Depending on available out off band information and motivated by the idea that mmWave is a location driven communication network, several mmWave BT schemes had been proposed in literature based on positioning information (PI). In this paper, we will make a survey on those works, clarifying how different positioning services contribute to reduce and relax BT complexity in mmWave communication. In this context, we will divide those studies into two main categories: (1) Straightforward positioning based BT schemes, which use PI for minimizing the searching space of a predefined codebook or reducing the complexity of channel estimation, (2) Positioning-based BT schemes using machine learning approaches, which classified as, first, statistical and probabilistic learning based schemes, secondly, other ML approaches based schemes. Moreover, we will discuss the effect of positioning and orientation errors, existence of obstacles, user mobility and storing information on the performance of BT. In addition, a comparison between all studies, will be presented considering several aspects such as the implementation cost. Finally, challenges facing these schemes will be discussed and several solutions will be suggested as possible future works.

High-Dimensional Feature Based Non-Coherent Detection for Multi-Intensity Modulated Ultraviolet Communications

Ultraviolet communication (UVC) has been regarded as a promising supplement for overloaded conventional wireless communications. One challenge lies in the communication deterioration caused by the UV-photon scattering induced inter-symbol-interference (ISI), which will be even worse when encountering multilevel pulse amplitude modulation (multi-PAM) symbols. To address this ISI, traditional coherent detection methods (e.g., maximum-likelihood sequence detection, MLSD) require high computational complexities for UV channel estimation and sequential detection space formation, thereby making them less attractive. Current non-coherent detection, which simply combines the ISI-insensitive UV signal features (e.g., the rising edge) into a one-dimensional (1D) metric, cannot guarantee reliable communication accuracy. In this work, a novel high-dimensional (HD) non-coherent detection scheme is proposed, leveraging a HD construction of the ISI-insensitive UV signal features. By doing so, we transform the ISI caused sequential detection into an ISI-released HD detection framework, which avoids complex channel estimation and sequential detection space computation. Then, to compute the detection surface, a UV feature based unsupervised learning approach is designed. We deduce the theoretical bit error rate (BER), and prove that the proposed HD non-coherent detection method has a lower BER than that of the current 1D non-coherent scheme. Simulation results validate our results, and more importantly, demonstrate a BER that approaches that of the state-of-the-art coherent MLSD ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$&lt; $</tex-math></inline-formula> 1 dB in SNR at BER = <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$4.5\times 10^{-3}$</tex-math></inline-formula> , the 7% overhead forward-error-correction limit), and also a reduction of computational complexity by at least two orders of magnitude.

Joint Power and User Grouping Optimization in Cell-Free Massive MIMO Systems

To relieve the stress on channel estimation and decoding complexity in cell-free massive multiple-input multiple-output (MIMO) systems, user grouping problem is investigated in this paper, where access points (APs) based on time-division duplex (TDD) are considered to serve users on different time resources and the same frequency resource. In addition, when quality of service (QoS) requirements are considered, widely-used max-min power control is no longer applicable. We derive the minimum power constraints under diverse QoS requirements considering user grouping. Based on the analysis, we formulate the joint power and user grouping problem under QoS constraints, aiming at minimizing the total transmit power. A generalized benders decomposition (GBD) based algorithm is proposed, where the primal problem and master problem are solved iteratively to approach the optimal solution. Simulation results demonstrate that by user grouping, the number of users served in cell-free MIMO systems can be as much as the number of APs without increasing the complexity of channel estimation and decoding. Furthermore, with the proposed user grouping strategy, the power consumption can be reduced by 2–3 dB compared with the reference user grouping strategy, and by 7 dB compared with the total transmit power without grouping.

Low Complex Methods for Robust Channel Estimation in Doubly Dispersive Environments

Wireless communications play a significant role in facilitating several mobile applications like unmanned aerial vehicles, high-speed railway, and vehicular communications. Particularly, the concept of connected vehicles brings a new level of connectivity to vehicles. Along with novel on- board computing and sensing technologies, vehicular networks serve as a key enabler of intelligent transportation systems and smart cities. However, in such environments, the propagation medium between the network nodes is highly time-varying leading to considerable reliability challenges. Ensuring communication reliability by the means of accurate channel estimation in such environments is very important. Initially, vehicular communications standards apply the basic least square (LS) estimation that is not enough for the dynamic vehicular environment. Moreover, the frame structure has low pilot density, making channel tracking a difficult task to achieve, especially in high mobility scenarios. Conventional estimators either employ data subcarriers besides pilots in the estimation process, or the estimated channel and noise statistics. Therefore they suffer from significant performance degradation due to high error probability resulting from hard symbol demapping and the sensitivity against the change in the employed channel statistics. The motivation behind this paper is to overcome this challenge by proposing a low complex and robust channel estimation scheme based on truncated discrete Fourier transform (T-DFT) that updates the channel estimates using DFT interpolation without the need for data subcarriers decisions and the estimated channel statistics. Moreover, further performance improvement can be achieved by considering temporal averaging on top of T-DFT estimation. Analytical and simulation results carried out using different vehicular channel models reveal the performance superiority of the proposed schemes compared to conventional estimators while recording a significant decrease in computational complexity and execution time.

Fast Fading Channel Estimation for OFDM Systems with Complexity Reduction

Channel estimation plays a significant role in the Inter-carrier interference (ISI) mitigation and symbol detection in Orthogonal frequency division multiplexing (OFDM) systems. As the speed of mobile terminal increases, the channel changes rapidly, causing fast fading which will degrade the system performance. In this paper, a low complexity fast fading channel estimation method suited for scattered pilot OFDM systems is proposed. This method is characterized by estimating channel matrix directly in the frequency domain, which can significantly reduce the channel estimation complexity with good system performance. Simulation results verify its superiority over the existing estimation methods.

Analytical Blind Beamforming for a Multi-Antenna UAV Base-Station Receiver in Millimeter-Wave Bands.

Over the last decade, unmanned aerial vehicles (UAVs) with antenna arrays have usually been employed for the enhancement of wireless communication in millimeter-wave bands. They are commonly used as aerial base stations and relay platforms in order to serve multiple users. Many beamforming methods for improving communication quality based on channel estimation have been proposed. However, these methods can be resource-intensive due to the complexity of channel estimation in practice. Thus, in this paper, we formulate an MIMO blind beamforming problem at the receivers for UAV-assisted communications in which channel estimation is omitted in order to save communication resources. We introduce one analytical method, which is called the analytical constant modulus algorithm (ACMA), in order to perform blind beamforming at the UAV base station; this relies only on data received by the antenna. The feature of the constant modulus (CM) is employed to restrict the target user signals. Algebraic operations, such as singular value decomposition (SVD), are applied to separate the user signal space from other interferences. The number of users in the region served by the UAV can be detected by exploring information in the measured data. We seek solutions that are expressible as one Kronecker product structure in the signal space; then, the beamformers that correspond to each user can be successfully estimated. The simulation results show that, by using this analytically derived blind method, the system can achieve good signal recovery accuracy, a reasonable system sum rate, and acceptable complexity.

OTFS underwater acoustic communications based on passive time reversal

Orthogonal time-frequency-space (OTFS) is a new two-dimensional (2D) modulation technique that offers reliable communications over time-varying channels. Due to the characteristics of underwater acoustic (UWA) channels, the computational complexity of traditional OTFS receivers is too high to implement in practice. In this paper, a 2D passive time reversal (PTR) structure receiver is designed in the Delay-Doppler domain, achieving spatial, temporal, and frequency focusing. Compared with existing methods, the proposed method could achieve similar performance with a lower complexity. Meanwhile, we also propose an adaptive channel estimation method based on the improved proportionate normalized least mean squares (IPNLMS) algorithm to reduce the complexity of channel estimation for the OTFS UWA communications system. The performance is evaluated under statistical simulated channels and measured channels from real-world experiments. The results verify the proposed 2D PTR receiver and demonstrate it could achieve good BER performance with a lower complexity.

A Modified OMP Algorithm with Reduced Feedback Overhead for Massive MIMO System

Objectives: Massive Multiple-Input and Multiple-Output (MIMO) is the optimum way to enhance the bandwidth issue, in which the feedback overhead is a challenging concern when tested with Frequency Division Duplex (FDD) systems. Conventional greedy search algorithms like Matching Pursuit (MP) and Orthogonal Matching Pursuit (OMP) are based on the prior knowledge of the signal sparsity, which enhances overhead. So, the objective of this study is to develop a method with low complexity channel estimation and reduced feedback overhead. Methods: The proposed method is based on the residual vector which is updated after iteration and a threshold-based approach is utilized to decide the stopping criteria thereby controlling the number of iterations. The evaluation parameters used in the experimental analysis are Normalized Mean Square Error and channel capacity. Findings: The proposed method improves the effectiveness in signal recovery and reconstruction along with signal sparsity it also reduces the complexity of the estimation process and limits the feedback. This results in a high data rate and can be used in a massive MIMO system to optimize the channel estimation. Novelty : The proposed model is independent of the sparsity in the input vector and improves the effectiveness of the signal recovery and reduces the feedback burden. The proposed method is suitable for a 5G massive MIMO systems. Keywords: Feedback overhead; CSI; OMP; Residue vector

ADMM Based Channel Estimation for RISs Aided Millimeter Wave Communications

In reconfigurable intelligent surfaces (RISs) aided millimeter wave communication systems, accurate channel state information (CSI) is particularly crucial to achieve high passive beamforming gain. However, the complexity of channel estimation increases significantly due to the high-dimensional channel matrix and the additional RISs' assisted channels. To tackle this challenge, an alternating direction method of multipliers (ADMM) based channel estimation method is proposed to improve the CSI estimation accuracy and reduce the training overhead. By converting the cascaded channel matrix into a sparse matrix recovery problem, both low rank and sparsity properties are exploited by the ADMM algorithm to solve the sparse matrix recovery problem. The simulation results verify that the proposed ADMM algorithm can significantly increase the accuracy of channel estimation and reduce the training overhead, compared with the well-known competing channel estimation algorithms.

Channel Estimation-Free Deep Direct Beamforming With Low Complexity in mmWave Massive MIMO

Hybrid precoding is used in millimeter wave (mmWave) massive multiple-input multiple-output (MIMO) systems instead of full digital precoding to reduce the cost and power consumption. Conventional hybrid precoding however requires complex channel estimation before beamforming. In this paper, we propose a low complexity low-cost channel estimation-free deep direct beamforming method composed of two basic steps-beam impulse and beamforming for time division duplex (TDD). This method is built upon randomized singular value decomposition (RSVD) and can directly complete the process of beamforming. Furthermore, a realistic dataset of mmWave channel-DeepMIMO is adopted to evaluate and determine the practicability of this method. The simulation results demonstrate that the proposed deep direct beamforming achieves high performance while its complexity and power consumption are very low. Compared with other methods without perfect channel status information (CSI), the proposed method costs less and requires lower computational resources while achieves higher spectral efficiency. Furthermore, the RSVD can be easily combined with conventional beamforming methods hence the proposed method is easily extendable to other applications.