E. Alameda-Hernandez

Channel estimation and symbol detection for block transmission using data-dependent superimposed training

We address the problem of frequency-selective channel estimation and symbol detection using superimposed training. The superimposed training consists of the sum of a known sequence and a data-dependent sequence that is unknown to the receiver. The data-dependent sequence cancels the effects of the unknown data on channel estimation. The performance of the proposed approach is shown to significantly outperform existing methods based on superimposed training (ST).

Frame/Training Sequence Synchronization and DC-Offset Removal for (Data-Dependent) Superimposed Training Based Channel Estimation

Over the last few years there has been growing interest in performing channel estimation via superimposed training (ST), where a training sequence is added to the information-bearing data, as opposed to being time-division multiplexed with it. Recent enhancements of ST are data-dependent ST (DDST), where an additional data-dependent training sequence is also added to the information-bearing signal, and semiblind approaches based on ST. In this paper, along with the channel estimation, we consider new algorithms for training sequence synchronization for both ST and DDST and block (or frame) synchronization (BS) for DDST (BS is not needed for ST). The synchronization algorithms are based on the structural properties of the vector containing the cyclic means of the channel output. In addition, we also consider removal of the unknown dc offset that can occur due to using first-order statistics with a non-ideal radio-frequency receiver. The subsequent bit error rate (BER) simulations (after equalization) show a performance not far removed from the ideal case of exact synchronization. While this is the first synchronization algorithm for DDST, our new approach for ST gives identical results to an existing ST synchronization method but with a reduced computational burden. In addition, we also present analysis of BER simulations for time-varying channels, different modulation schemes, and traditional time-division multiplexed training. Finally, the advantage of DDST over (conventional, non semi-blind) ST will reduce as the constellation size increases, and we also show that even without a BS algorithm, DDST is still superior to conventional ST. However, iterative semiblind schemes based upon ST outperform DDST but at the expense of greater complexity

The Averaged, Overdetermined, and Generalized LMS Algorithm

This paper provides and exploits one possible formal framework in which to compare and contrast the two most important families of adaptive algorithms: the least-mean square (LMS) family and the recursive least squares (RLS) family. Existing and well-known algorithms, belonging to any of these two families, like the LMS algorithm and the RLS algorithm, have a natural position within the proposed formal framework. The proposed formal framework also includes - among others - the LMS/overdetermined recursive instrumental variable (ORIV) algorithm and the generalized LMS (GLMS) algorithm, which is an instrumental variable (IV) enable LMS algorithm. Furthermore, this formal framework allows a straightforward derivation of new algorithms, with enhanced properties respect to the existing ones: specifically, the ORIV algorithm is exported to the LMS family, resulting in the derivation of the averaged, overdetermined, and generalized LMS (AOGLMS) algorithm, an overdetermined LMS algorithm able to work with an IV. The proposed AOGLMS algorithm overcomes - as we analytically show here - the limitations of GLMS and possesses a much lower computational burden than LMS/ORIV, being in this way a better alternative to both algorithms. Simulations verify the analysis.

A fourth-order stationarity and ergodicity conditions for harmonic processes

This paper considers the problem of estimating the fourth-order cumulant sequence of a harmonic process. Both stationarity and ergodicity conditions for this kind of signal are derived through the imposition of restrictions to the frequencies and amplitudes that define the signals. Statistical properties for the mean value and the variance of the sample mean estimator are deduced as well. These conditions are applied to cubic-phase coupling detection, stressing the study of nonergodic signals and possible strategies to study them. Numerical examples to illustrate these conditions for cubic phase coupling detection and separation of mixed signals using independent component analysis (ICA) methods are also showed in this paper.

Improved synchronisation for superimposed training based channel estimation

This paper introduces a synchronisation method for super-imposed training (ST) based channel estimation, using periodic ST sequences. The method exploits the particular structure, occurring when the ST sequence period is larger than the channel length, of the vector containing the received signals first-order cyclostationary statistics. After synchronisation, any DC-offset can be removed and an unbiased channel estimate can be obtained. Necessary and sufficient conditions for synchronisation are provided. The problem of training sequence design for an improved synchronisation is also addressed. An expression for the variance of the channel estimate is obtained as well, assuming perfect synchronisation and using the designed training sequences. The proposed synchronisation method is computationally more efficient than existing methods, and yet its performance, in term of channel estimation MSE and BER, is not diminished as shown by simulations

Implicitly-trained channel estimation and equalization with zero mean input data packets

Implicit training (IT) channel estimation adds a periodic training sequence to each input data block/packet, so that no bandwidth is lost (as in a traditionally trained scenario). While the input data is usually assumed to be zero mean, each data packet will have a deterministic mean, which is itself a random variable. In this paper we show that by removing this nonzero mean for each input packet before transmission and then employing the IT method, we improve the channel estimate, when compared to the normal IT approach. In addition, if we then implement a MMSE equalizer (based upon the improved channel estimate), the BER is also improved (even with nonzero mean removal of each packet) when compared to MMSE equalization based on the traditional IT channel estimation.

A new method for blind identification of FIR channels based almost exclusively on second order statistics

Here we show that by using second-order statistics (SOS) of a system output all the spectrally equivalent (SE) systems can be derived, and the correct phase can then be chosen by applying appropriately non-linear techniques such as cumulants. In this paper, a new method is proposed that first extracts higher-order statistics (HOS) information (i.e. phase) and then computes directly the true system using only SOS, thus avoiding searching the whole SE set. This new approach depends minimally on HOS, thus avoiding the well-understood limitations of HOS calculations. Finally, comparisons are made with similar approaches through both simulation and analysis, and the advantages of this new method are documented.