The discrete Fourier transform (DFT) precoded orthogonal frequency-division multiplexing (OFDM) has been adopted as the uplink transmission technique in the long-term evolution terrestrial communication standard, for its lower peak-to-average power ratio (PAPR) and similar receiver complexity, compared with the standard OFDM. However, its application in the underwater acoustic (UWA) communications remains doubtful, mainly for the lack of systematic studies as well as sufficient experimental verifications. This paper provides a comprehensive investigation on the DFT-precoded OFDM UWA communication with a multiple-input-multiple-output (MIMO) configuration, and corroborates its superiority for implementing a practical UWA communication modem. The DFT precoding is applied on the data symbols to achieve a lower PAPR than that of the standard OFDM, and it is optional for the pilot symbols. The frequency-domain turbo equalization (FDTE) technique, especially suitable for interference-intensive scenarios, is employed on the receiver side to combat the intersymbol interference (ISI) and the multiplexing interference. Experimental results are provided to demonstrate the performance of the proposed transceiver scheme. It is shown reliable communication is achieved for the two-transducer transmission with a QPSK modulation and the one-transducer transmission with a 16QAM modulation, even without running any iteration for the FDTE. With the help of turbo iterations, a two-transducer transmission with a 16QAM modulation also achieves a satisfactory performance.

The proportionate updating (PU) mechanism has been widely adopted in least mean squares (LMS) adaptive filtering algorithms to exploit the system sparsity. In this paper, we propose a proportionate recursive least squares (PRLS) algorithm for the sparse system estimation, in which, an independent weight update is assigned to each tap according to the magnitude of that estimated filter coefficient. Its mean square performance is analyzed via the energy conservation principle in both the transient and steady-state stages. In this way, an explicit condition on the control parameter of the proportionate matrix of PRLS can be obtained to ensure a better steady-state performance than that of RLS. Simulation results in a system identification setting support the analysis.

Nowadays, with the development of automotive and traffic radars, a higher angle resolution is required for an increasing demand of four-dimensional (4D) imaging radar. In this paper, the high-precision direction-of-arrival (DOA) estimation problem is solved by using a deep learning (DL) framework. Most existing on-grid DL-based methods have an upper limit of one-degree resolution. The DOA estimation performance under such resolution is still far behind conventional methods, and is not accurate enough in practical applications. Hence, we introduce a two-stage multi-layer perceptron (TS-MLP) framework to achieve higher resolution of DOA estimation with low complexity by dividing the problem into two main parts. The first MLP is used to determine the coarse grid point nearest to the source angle, and the second MLP fine tunes the estimate within the coarse grids. In addition, we propose a solution to the source association problem between the two stages when handling multiple targets. Our scheme shows much higher accuracy compared with existing DL-based methods, and has comparable performance with traditional high-resolution methods. Moreover, it performs quite robust in the presence of array imperfections. The demo code is available in https://github.com/Whisperzyj/TS-MLP

In order to play the key role of automotive millimeter wave radar in intelligent vehicle systems, direction-of-arrival (DOA) estimation is an essential problem to be solved. For practical intelligent driving applications, DOA estimation requires both real-time performance and high accuracy. Due to unique advantages, deep learning (DL) based methods have attracted more attention. Most of the existing DL-based methods require a large number of snapshots, but only a few snapshots can be guaranteed in practical applications. Moreover, they usually model DOA estimation as a multi-label classification task. The output represents the position of signal DOA on the discrete grid, and the resolution will be limited by the grid. In this paper, a new space-time Network (STNet) is proposed, which models DOA estimation as a regression task to achieve the effect of gridless estimation. We design a space correlation extraction module (SCEM) and a time correlation extraction module (TCEM), using the covariance matrix of the received signal and the original received signal as inputs respectively, treat them as different types of data. In these two modules, skip connection dense blocks (SCDBs) and long short-term memory (LSTM) networks are adopted to process two different forms of data. Through such processing, we retain sufficient information, obtain more features for the regression task, and ensure the estimation effect of using a small number of snapshots. The experimental results indicate that the STNet shows obvious performance gain in the case of small snapshots, achieves gridless estimation effect, and demonstrates excellent adaptability in situations where target DOAs are closely positioned.