This paper studies the mean and mean-square behaviors of the M-estimate based normalized subband adaptive filter algorithm (M-NSAF) with robustness against impulsive noise. Based on the contaminated-Gaussian noise model, the stability condition, transient and steady-state results for the algorithm are formulated analytically. The analysis results help us to better understand the M-NSAF performance in impulsive noise. To further obtain fast convergence and low steady-state estimation error, we derive a variable step size (VSS) M-NSAF algorithm. This VSS scheme is also generalized to proportionate M-NSAF variants for sparse systems. Computer simulations on system identification in impulsive noise and acoustic echo cancellation with double-talk are performed to demonstrate our theoretical analysis and the effectiveness of the proposed algorithms.

Multi-input multi-output (MIMO) detection based on turbo principle has been shown to provide a great enhancement in the throughput and reliability of underwater acoustic (UWA) communication systems. Benefits of the iterative detection in MIMO systems, however, can be obtained only when a high quality channel estimation is ensured. In this paper, we develop a new soft-decision-driven sparse channel estimation and turbo equalization scheme in the triply selective MIMO UWA. First, the Homotopy recursive least square dichotomous coordinate descent (Homotopy RLS-DCD) adaptive algorithm, recently proposed for sparse single-input single-output (SISO) system identification, is extended to adaptively estimate rapid time-varying MIMO sparse channels. Next, the more reliable a posteriori soft-decision symbols, instead of the hard decision symbols or the a priori softdecision symbols, at the equalizer output, are not only feedback to the Homotopy RLS-DCD based channel estimator but also to the minimum mean-square-error (MMSE) equalizer. As the turbo iterations progress, the accuracy of channel estimation and the quality of the MMSE equalizer are improved gradually, leading to the enhancement in the turbo equalization performance. This also allows the reduction in pilot overhead. The proposed receiver has been tested by using the data collected from the SHLake2013 experiment. The performance of the receiver is evaluated for various modulation schemes, channel estimators and MIMO sizes. Experimental results demonstrate that the proposed a posteriori soft-decision-driven sparse channel estimation based on the Homotopy RLS-DCD algorithm and turbo equalization offer considerable improvement in system performance over other turbo equalization schemes.

Receivers with joint channel estimation (CE) and signal detection using superimposed pilots (SP) can achieve high transmission efficiency in orthogonal time frequency space (OTFS) systems. However, existing receivers have high computational complexity, hindering their practical applications. In this work, with SP in the delay-Doppler (DD) domain and the generalized complex exponential (GCE) basis expansion modeling (BEM) for channels, a message passing-based SP-DD iterative receiver is proposed, which drastically reduces the computational complexity while with marginal performance loss, compared to existing ones. To facilitate CE in the proposed receiver, we design pilot signal to achieve pilot power concentration in the frequency domain, thereby developing an SP-DD-D receiver that can reduce the power of the pilot signal with almost no loss of CE and bit error rate (BER) performance. Extensive simulation results are provided to demonstrate the superiority of the proposed SP-DD-D receiver.

This article presents the power of deep learning for signal detection in orthogonal frequency division multiplexing with index modulation (OFDM-IM) systems. In this article, we propose the Y-shaped net-based scheme with fully connected layers (Y-FC) and with bidirectional long short-term memory units (Y-BLSTM) to further optimize the data reception. In the Y-shaped scheme, the active indices and the constellation demapping are learned by two parallel sub-neural networks respectively, the output features are then concatenated and flattened to obtain binary bits. The performance is evaluated under various system parameters, such as modulation methods, number of hidden neural nodes and channel states. Simulation results show that compared with the data-driven signal detection with direct input-output, the proposed scheme achieves several dB of bit-error-rate performance improvement with almost the same running time.

In this paper, a novel kernel mixed error criterion (KMEC) algorithm is proposed for nonlinear system identification, which uses a combination of two different error schemes to implement a newly constructed cost function, which is realized by using a logarithmic squared error and a generalized maximum correntropy criterion (GMCC) to devise the KMEC algorithm. The proposed KMEC is derived in the context of the kernel adaptive filter and it provides good performance for identifying the nonlinear channels in different mixed noise environments in terms of the mean square error (MSE) at its steady-state and convergence performance.