Showing papers by "Silviu Ciochina published in 2017"

On the Identification of Bilinear Forms With the Wiener Filter

Abstract: In this letter, the identification problem of bilinear forms with the Wiener filter is addressed. The contribution is twofold. First, a different approach is introduced, by defining the bilinear term with respect to the impulse responses of a spatiotemporal model, in the context of multiple-input/single-output systems. Second, two versions of the Wiener filter (namely direct and iterative) are developed in this context. Moreover, the advantage of the iterative Wiener filter is outlined as compared to the direct solution. The results of the simulations, which are performed from a system identification perspective, support the theoretical findings.

Robust variable-regularized RLS algorithms

Abstract: The regularization parameter is required in most (if not all) adaptive algorithms, while its role becomes very critical in the presence of additive noise. In this paper, we focus on the regularized recursive least-squares (RLS) algorithm and present a method to find its regularization parameter, which is related to the signal-to-noise ratio (SNR). Also, using a proper estimation of the SNR, we further propose a variable-regularized RLS (VR-RLS) algorithm. In addition, a low-complexity version of the VR-RLS algorithm is developed, based on the dichotomous coordinate descent (DCD) method. Due to their nature, the proposed algorithms have good robustness features against additive noise, which make them behave well in all noisy conditions. Simulations performed in the context of acoustic echo cancellation support these findings.

An NLMS algorithm for the identification of bilinear forms

Abstract: In a recent work, we addressed the identification poblem of bilinear forms with the Wiener filter. In this context, a different approach was introduced, by defining the bilinear term with respect to the impulse responses of a spatiotemporal model, which resembles a multiple-input/single-output (MISO) system. However, in practice, the Wiener filter may not be always very efficient or convenient to use. Consequently, in this paper, we further develop a normalized least-mean-square (NLMS) adaptive filter tailored for bilinear forms. Many simulations, which are performed from a system identification perspective, indicate the good performance of the proposed algorithm.

Analysis of an LMS algorithm for bilinear forms

Abstract: In this paper, we present a least-mean-square (LMS) adaptive filter tailored for bilinear forms. The bilinear term is defined with respect to the impulse responses of a spatiotemporal model, which resembles a multiple-input/single-output (MISO) system. A convergence analysis is also provided, outlining the main features of this algorithm. Simulation results support the theoretical findings.

A reduced complexity multi-user massive MIMO precoding method

Abstract: Considering the upcoming future generation of mobile communication standards, Multi-User Massive MIMO (MU-MMIMO) is a well-known method of increasing the overall cell throughput. Such a scenario suffers from a high computational complexity because of the transmit precoding that is used to eliminate the inter-user interference. The paper proposes a reduced complexity precoding method applicable at the base station (BS) side of a downlink multiuser scenario. The reduction in complexity is achieved by reducing the number of low-level operations involved in the QR decomposition without any effect on the overall system performance. The method is shown to be compatible with two of the most widespread implementations of the QR decomposition: Givens rotations and Householder reflections. The proposed algorithm is evaluated in terms of required number of multiply-and-accumulate (MAC) operations and compared with its unmodified version, showing a net decrease of computational complexity.

Investigation on the performances of APA in forensic noise reduction

Abstract: Multimedia files, either video or audio, could greatly influence the final verdict of a trial when accepted as evidence. The abundance of free editing software available nowadays make forgeries a very easy operation. Audio messages, even if authentic, in some cases, can be heavily masked by other signals and declared unusable. This paper presents the investigations on the performance of the affine projection algorithm (APA) in recovering a speech signal drowned in a loud musical signal and it represents a contribution to the multimedia forensic domain. The APA was tested in multiple situations showing the top performance limits and how the working parameters influence those limits.

An RLS algorithm for the identification of bilinear forms

Abstract: Bilinear systems are involved in many interesting applications, especially related to the approximation of nonlinear systems. In this context, the bilinear term is usually defined in terms of an input-output relation (i.e., with respect to the data). Recently, a different approach has been introduced, by defining the bilinear term with respect to the impulse responses of a spatiotemporal model, which resembles a multiple-input/single-output (MISO) system. Also, in this framework, the Wiener filter has been involved to address the identification problem of these bilinear forms. Since the Wiener filter may not be always very efficient or convenient to use in practice, we propose in this paper an adaptive filtering approach based on the recursive least-squares (RLS) algorithm. Simulations performed in the context of system identification (based on the MISO system approach) indicate the appealing performance of this algorithm.

An adaptive solution for nonlinear system identification

Abstract: This paper proposes a new and very simple method to perform identification of nonlinearities in systems that are supposed to be linear. The improvement with respect to the previous approaches lies in the low computational complexity of the solution and the convenient possibility of online evaluation. Simulation results show that the proposed method provides good performance in terms of coefficients identification, resulting in both low steady-state error and high convergence rate of the algorithm, which justifies its applicability in practice.

Optimized block-diagonalization precoding technique using givens rotations QR decomposition

Abstract: The emerging 5G mobile communication standard aims to increase the throughput and, in the same time, to considerably increase the number of users serviced concurrently (Internet of Things). The key direction for achieving these requirements is to heavily extend the use of the spatial degrees of freedom, especially in the multi-user scenarios. Multi-User Massive MIMO (MU-MIMO) is one of the key technologies that responds well to the 5G needs. However, the use of MU-MIMO in the downlink direction raises the collaborative detection problem at the user side, thus the elimination of the inter-user interference becomes necessary. The paper presents a reduced complexity linear transmitter precoding technique that cancels the inter-user interference in a downlink MU-MIMO system. The reduced complexity is achieved through re-using as many low-level operations as possible. The method is suitable for implementation on any modern processor and proven to be scalable to a Massive MIMO scenario without any loss in performance.

Improved convergence model of the affine projection algorithm for system identification

Abstract: The convergence analysis of the affine projection algorithm (APA) was intensively studied, however, it is far to be a trivial task. In this context, a set of assumptions have to be considered, in order to allow a tractable analytical approach. As a consequence, notable differences occur between the simulation results and the analytical model. In this paper, the convergence of APA is studied and a refined approximation is proposed, named “the third level of approximation.” The results show a near perfect matching between the theoretical model and simulation results, in terms of the asymptotic behavior and convergence speed of the studied algorithm.