We propose a new iterative detection and decoding (IDD) algorithm for multiple-input multiple-output (MIMO) based on expectation propagation (EP) with application to massive MIMO scenarios. Two main results are presented. We first introduce EP to iteratively improve the Gaussian approximations of both the estimation of the posterior by the MIMO detector and the soft output of the channel decoder. With this novel approach, denoted by double-EP (DEP), the convergence is very much improved with a computational complexity just two times the one of the linear minimum mean square error (LMMSE) based IDD, as illustrated by the included experiments. Besides, as in the LMMSE MIMO detector, when the number of antennas increases, the computational cost of the matrix inversion operation required by the DEP becomes unaffordable. In this work we also develop approaches of DEP where the mean and the covariance matrix of the posterior are approximated by using the Gauss-Seidel and Neumann series methods, respectively. This low-complexity DEP detector has quadratic complexity in the number of antennas, as the low-complexity LMMSE techniques. Experimental results show that the new low-complexity DEP achieves the performance of the DEP as the ratio between the number of transmitting and receiving antennas decreases.

This paper focuses on the offline handwriting text recognition problem (HTR) with small training data sets. Some techniques such as transfer learning or data augmentation have recently been applied to this problem, improving the performance of the recognition. In these scenarios, we found that errors in the labelling of the training samples, present in some databases, have a great impact in the character error rates (CER). Accordingly, we propose a novel cross validation technique to remove incorrect labelled lines. In this approach, after a first training stage, transcript lines with CER above a threshold are discarded, where the threshold is a function of the available data. Less available data favours larger CER, even for healthy lines, suggesting higher thresholds for fewer lines. This new technique and the validation of the threshold are analyzed over the ICFHR 2018 competition on automated HTR and other well known databases such as Washington and Parzival. For the Ricordi database in the ICFHR 2018, with transcription errors, we report a reduction of CER by 2%.

In this paper we propose a novel turbo equalizer based on the expectation propagation (EP) algorithm. Optimal equalization is computationally unfeasible when high-order modulations and/or large memory channels are used. In these scenarios, low-cost and suboptimal equalizers, such as those based on the linear minimum mean square error (LMMSE), are commonly used. The LMMSE-based equalizer can be efficiently implemented with a Kalman smoother (KS), i.e., a forward and backward Kalman filtering whose predictions are merged in a posterior smoothing step. Recently, it was shown that applying EP at the forward and backward stages of a KS equalizer could significantly improve its performance. In this paper, we investigate applying EP at the smoothing level instead. Also, we propose some further modifications to better exploit the information coming from the channel decoder in turbo equalization schemes. Overall, we remarkably reduce the computational complexity while highly improving the performance in terms of bit error rate.

In this paper we deal with the offline handwriting text recognition (HTR) problem with reduced training data sets. Recent HTR solutions based on artificial neural networks exhibit remarkable solutions in referenced databases. These deep learning neural networks are composed of both convolutional (CNN) and long short-term memory recurrent units (LSTM). In addition, connectionist temporal classification (CTC) is the key to avoid segmentation at character level, greatly facilitating the labeling task. One of the main drawbacks of the CNN-LSTM-CTC (CRNN) solutions is that they need a considerable part of the text to be transcribed for every type of calligraphy, typically in the order of a few thousands of lines. Furthermore, in some scenarios the text to transcribe is not that long, e.g. in the Washington database. The CRNN typically overfits for this reduced number of training samples. Our proposal is based on the transfer learning (TL) from the parameters learned with a bigger database. We first investigate, for a reduced and fixed number of training samples, 350 lines, how the learning from a large database, the IAM, can be transferred to the learning of the CRNN of a reduced database, Washington. We focus on which layers of the network could not be re-trained. We conclude that the best solution is to re-train the whole CRNN parameters initialized to the values obtained after the training of the CRNN from the larger database. We also investigate results when the training size is further reduced. For the sake of comparison, we study the character error rate (CER) with no dictionary or any language modeling technique. The differences in the CER are more remarkable when training with just 350 lines, a CER of 3.3% is achieved with TL while we have a CER of 18.2% when training from scratch. As a byproduct, the learning times are quite reduced. Similar good results are obtained from the Parzival database when trained with this reduced number of lines and this new approach.