Tesis (Teoría de la Señal y Comunicaciones)

URI permanente para esta colecciónhttps://hdl.handle.net/11441/11425

Examinar

Mostrando 1 - 3 de 3

Expectation propagation as a solution for digital communication systems.
(2018-11-16) Santos Velázquez, Irene; Arias de Reyna Domínguez, Eva María; Murillo Fuentes, Juan José; Universidad de Sevilla. Departamento de Teoría de la Señal y Comunicaciones
In the context of digital communications, a digital receiver is required to provide an estimation of the transmitted symbols. Nowadays channel decoders highly benefit from soft (probabilistic) estimates for each transmitted symbol rather than from hard decisions. For this reason, digital receivers must be designed to provide the probability that each possible symbol was transmitted based on the received corrupted signal. Since exact inference might be unfeasible in terms of complexity for high-order scenarios, it is necessary to resort to approximate inference, such as the linear minimum mean square error (LMMSE) criterion. The LMMSE approximates the discrete prior information of the transmitted symbols with a Gaussian distribution, which causes a degradation in its performance. In this thesis, an alternative approximate statistical technique is applied to the design of a digital probabilistic receiver in digital communications. Specifically, the expectation propagation (EP) algorithm is investigated to find the Gaussian posterior probability density function (pdf) that minimizes the Kullback-Leibler (KL) divergence with respect to the true posterior pdf. Two different communication system scenarios are studied: a single-input singleoutput (SISO) digital communication system with memory channel and a multipleinput multiple-output (MIMO) system with memoryless channel. In the SISO scenario, three different designs of a soft standalone and turbo equalizer based on the EP algorithm are developed: the block or batch approach, the filter-type version that emulates theWiener filter behavior and the smoothing equalizer which proceeds similarly to a Kalman smoother. Finally, the block EP implementation is also adapted to MIMO scenarios with feedback from the decoder. In both scenarios, the EP is applied iteratively, including a damping mechanism and a control to avoid negative values of variances, which would lead to instabilities (specially for high-order constellations). Experimental results included through the thesis show that the EP algorithm applied to communication systems greatly improves the performance of previous approaches found in the literature with a complexity slightly increased but still proportional to that of the LMMSE. These results also show the robustness of the algorithm even for high-order modulations, large memory channels and high number of antennas. Major contributions of this dissertation have been published in four journal (one of them is still under review) and two conference papers. One more paper will be submitted to a journal soon. All these papers are listed below: • Irene Santos, Juan José Murillo-Fuentes, Rafael Boloix-Tortosa, Eva Arias de Reyna and Pablo M. Olmos, "Expectation Propagation as Turbo Equalizer in ISI Channels," IEEE Transactions on Communications, vol. 65, no.1, pp. 360-370, Jan 2017. • Irene Santos, Juan José Murillo-Fuentes, Eva Arias de Reyna and Pablo M. Olmos, "Turbo EP-based Equalization: a Filter-Type Implementation," IEEE Transactions on Communications, Sep 2017, Accepted. [Online] Available: https://ieeexplore.ieee.org/document/8353388/ • Irene Santos, Juan José Murillo-Fuentes, Eva Arias-de-Reyna and Pablo M. Olmos, "Probabilistic Equalization With a Smoothing Expectation Propagation Approach," IEEE Transactions on Wireless Communications, vol. 16, no. 5, pp. 2950-2962, May 2017. • Irene Santos, Juan José Murillo-Fuentes and Eva Arias-de-Reyna, "Equalization with Expectation Propagation at Smoothing Level," To be submitted. [Online] Available: https://arxiv.org/abs/1809.00806 • Irene Santos and Juan José Murillo-Fuentes, "EP-based turbo detection for MIMO receivers and large-scale systems," IEEE Transactions on Vehicular Technology, May 2018, Under review. [Online] Available: https://arxiv.org/abs/1805.05065 • Irene Santos, Juan José Murillo-Fuentes, and Pablo M. Olmos, "Block expectation propagation equalization for ISI channels," 23rd European Signal Processing Conference (EUSIPCO 2015), Nice, 2015, pp. 379-383. • Irene Santos, and Juan José Murillo-Fuentes, "Improved probabilistic EPbased receiver for MIMO systems and high-order modulations," XXXIII Simposium Nacional de la Unión Científica Internacional de Radio (URSI 2018), Granada, 2018.
Machine Learning for handwriting text recognition in historical documents
(2021-12-17) Aradillas Jaramillo, Jose Carlos; Murillo Fuentes, Juan José; Martínez Olmos, Pablo; Universidad de Sevilla. Departamento de Teoría de la Señal y Comunicaciones
Olmos ABSTRACT In this thesis, we focus on the handwriting text recognition task over historical documents that are difficult to read for any person that is not an expert in ancient languages and writing style. We aim to take advantage and improve the neural networks architectures and techniques that other authors are proposing for handwriting text recognition in modern handwritten documents. These models perform this task very precisely when a large amount of data is available. However, the low availability of labeled data is a widespread problem in historical documents. The type of writing is singular, and it is pretty expensive to hire an expert to transcribe a large number of pages. After investigating and analyzing the state-of-the-art, we propose the efficient application of methods such as transfer learning and data augmentation. We also contribute an algorithm for purging mislabeled samples that affect the learning of models. Finally, we develop a variational auto encoder method for generating synthetic samples of handwritten text images for data augmentation. Experiments are performed on various historical handwritten text databases to validate the performance of the proposed algorithms. The various included analyses focus on the evolution of the character and word error rate (CER and WER) as we increase the training dataset. One of the most important results is the participation in a contest for transcription of historical handwritten text. The organizers provided us with a dataset of documents to train the model, then just a few labeled pages of 5 new documents were handled to adjust the solution further. Finally, the transcription of nonlabeled images was requested to evaluate the algorithm. Our method raked second in this contest.
Separación ciega de fuentes y su aplicación a receptores de comunicaciones digitales: CDMA, OFDM y MC-CDMA
(2005) Boloix Tortosa, Rafael; Murillo Fuentes, Juan José; Universidad de Sevilla. Departamento de Teoría de la Señal y Comunicaciones
La Tesis está dividida en dos partes. En la primera parte se realiza un estudio de las técnicas de separación ciega de fuentes (BSS), mientras que en la segunda se estudia la aplicación de tales técnicas a los sistemas de comunicaciones digitales, en concreto a CDMA, OFDM, y MC-CDMA.

Examinar

Examinando Tesis (Teoría de la Señal y Comunicaciones) por Autor "Murillo Fuentes, Juan José"