Fully adaptive gaussian mixture metropolis-hastings algorithm

Markov Chain Monte Carlo methods are widely used in signal processing and communications for statistical inference and stochastic optimization. In this work, we introduce an efficient adaptive Metropolis-Hastings algorithm to draw samples from generic multimodal and multidimensional target distributions. The proposal density is a mixture of Gaussian densities with all parameters (weights, mean vectors and covariance matrices) updated using all the previously generated samples applying simple recursive rules. Numerical results for the one and two-dimensional cases are provided.

Cross-Products LASSO

Negative co-occurrence is a common phenomenon in many signal processing applications. In some cases the signals involved are sparse, and this information can be exploited to recover them. In this paper, we present a sparse learning approach that explicitly takes into account negative co-occurrence. This is achieved by adding a novel penalty term to the LASSO cost function based on the cross-products between the reconstruction coefficients. Although the resulting optimization problem is non-convex, we develop a new and efficient method for solving it based on successive convex approximations. Results on synthetic data, for both complete and overcomplete dictionaries, are provided to validate the proposed approach.

Efficient Monte Carlo optimization for multi-label classifier chains

Multi-label classification (MLC) is the supervised learning problem where an instance may be associated with multiple labels. Modeling dependencies between labels allows MLC methods to improve their performance at the expense of an increased computational cost. In this paper we focus on the classifier chains (CC) approach for modeling dependencies. On the one hand, the original CC algorithm makes a greedy approximation, and is fast but tends to propagate errors down the chain. On the other hand, a recent Bayes-optimal method improves the performance, but is computationally intractable in practice. Here we present a novel double-Monte Carlo scheme (M2CC), both for finding a good chain sequence and performing efficient inference. The M2CC algorithm remains tractable for high-dimensional data sets and obtains the best overall accuracy, as shown on several real data sets with input dimension as high as 1449 and up to 103 labels.

Diseño y análisis de unidades didácticas para Procesado Digital de la Señal.

Este Proyecto Fin de Carrera pretende desarrollar una serie de unidades didácticas orientadas a mejorar el aprendizaje de la teoría de procesado digital de señales a través de la aplicación práctica. Con tal fin, se han diseñado una serie de prácticas que permitan al alumno alcanzar un apropiado nivel de conocimiento de la asignatura, la adquisición de competencias y alcanzar los resultados de aprendizaje previstos. Para desarrollar el proyecto primero se ha realizado una selección apropiada de los contenidos de la teoría de procesado digital de señales en relación con los resultados de aprendizaje esperados, seguidamente se han diseñado y validado unas prácticas basadas en un entorno de trabajo basado en MATLAB y DSP, y por último se ha redactado un manual de laboratorio que combina una parte teórica con su práctica correspondiente. El objetivo perseguido con la realización de estas prácticas es alcanzar un equilibrio teórico/práctico que permita sacar el máximo rendimiento de la asignatura desde el laboratorio, trabajando principalmente con el IDE Code Composer Studio junto con un kit de desarrollo basado en un DSP. ABSTRACT. This dissertation intends to develop some lessons oriented to improve about the digital signal processing theory. In order to get this objective some practices have been developed to allow to the students to achieve an appropriate level of knowledge of the subject, acquire skills and achieve the intended learning outcomes. To develop the project firstly it has been made an appropriate selection of the contents of the digital signal processing theory related with the expected results. After that, five practices based in a work environment based on Matlab and DSP have been designed and validated, and finally a laboratory manual has been drafted that combines the theoretical part with its corresponding practice. The objective with the implementation of these practices is to achieve a theoretical / practical balance to get the highest performance to the subject from the laboratory working mainly with the Code Composer Studio IDE together a development kit based on DSP.

Desarrollo de herramientas para la autoevaluación del Laboratorio de Procesado Digital de la Señal

Este proyecto tiene como objetivo el desarrollo de una herramienta que permita al alumno la autocorrección de prácticas de la asignatura de Procesado Digital de la Señal. La herramienta será diseñada por medio del GUI de Matlab, que permite la creación de interfaces gráficos de usuario para la interacción con el alumno, así él mismo podrá comprobar si los resultado obtenidos para el enunciado de la práctica facilitado son correctos. La evaluación del alumno se llevará a cabo pidiendo distintas respuestas sobre las prácticas y comparándolas posteriormente con los resultados correctos. El código invisible al usuario será el encargado de indicar si el resultado es correcto o no lo es. ABSTRACT. The aim of this project is to develop a tool for the students of Digital Signal Processing that help them self-correct their lab exercises. The tool will be designed using the Matlab GUI, which allows the creation of graphical user interfaces to interact with the student, who can check whether the results obtained are correct or not. The student will be asked about different results of the exercises and the answers will be compared with the correct results. A part of the tool hidden to the student will reveal to the lecturer the outcome of this comparison.

Bases para un diseño metodológico de estimulación binaural como terapia acústica en neurología

La medicina y la ingeniería del siglo XXI han dado como fruto numerosos avances para la sociedad aunque en la mayoría de los casos los tratamientos suelen ser costosos e invasivos. La educación que recibe la sociedad sobre la salud es escasa, ya que sólo vamos al médico cuando realmente estamos enfermos. Este trabajo presenta nuestra apuesta por las terapias complementarias, para el desarrollo de una metodología terapéutica no invasiva y con un costo muy bajo. La finalidad de esta Tesis, que se enmarca en un equipo multidisciplinar, fruto de la estrecha colaboración en el que participan psicopedagogos, ingenieros y médicos, es perfilar una metodología que luego pueda ser aplicable a patologías neurológicas. Aquí, dejamos sentadas las bases. Faltarán nuevos investigadores que continúen este camino para tener una base de datos lo suficientemente extensa de registros de sujetos que hayan sido sometidos a terapia binaural, para poder sacar unas conclusiones sólidas. La aportación de esta Tesis deja cubierta la aplicación, selección, procesado de señal y desarrollo de algoritmos, test cognitivos indicados para el caso específico que nos ocupa, cálculo de incertidumbre del sistema utilizado para la aplicación del estímulo y desarrollo de un test psicoacústico específico. EL empleo del sonido en medicina como es la musicoterapia o sonoterapia ha experimentado una gran difusión en los últimos años, más de 100.000 nuevas citas bibliográficas han aparecido con respecto al año anterior. Sin embargo, son escasísimas las que hacen referencia a las características físico acústicas del sonido empleado, tan sólo hemos encontrado una par de ellas que correlacionan las características físicas del sonido con el tipo de respuesta terapéutica. No encontramos citas bibliográficas específicas que planteen un modelo experimental científico capaz de reproducir las mismas respuestas ante los mismos parámetros y estímulos. En esta Tesis proponemos el uso de estimulación sonora binaural, que consiste en la utilización de dos tonos puros idénticos pero ligeramente diferentes en frecuencia que se presentan de manera separada cada uno en un oído, como consecuencia, la persona que recibe la estimulación percibe un tercer tono, llamado tono binaural, formado por la diferencia de frecuencia de ambos variando su amplitud. Existen estudios que sugieren que dichas frecuencias binaurales pueden modificar los patrones eléctricos de la actividad cerebral y los niveles de arousal, conociéndose en la literatura bajo el nombre de “entrainment”. Tras la revisión bibliográfica del estado del arte, podemos concluir que es necesario el desarrollo de estudios doble ciego bien diseñados, con el objetivo de establecer una base sólida sobre los efectos de este tipo de estimulación, ya que la mayoría de los beneficios documentados se refieren a muestras muy pequeñas y con poco rigor científico, siendo los resultados positivos obtenidos debidos al efecto placebo. La tecnología binaural es barata siendo cualquier avance en esta dirección de interés público. El objetivo concreto de la investigación es estudiar el potencial de las ondas binaurales en un área en particular: tareas que requieren atención y concentración. Se busca obtener cualquier cambio en las ondas cerebrales que se puedan correlar con la mejoras. A la vista de los resultados de estas investigaciones se intentará aplicar esta metodología en neuropatologías que presenten alguna deficiencia en el área de atención como es el Trastorno de espectro Autista. En esta Tesis presentamos los resultados de dos estudios independientes, el primero para sentar las bases del método (tiempos, diseño de estimulaciones, procesado) en una muestra de 78 adultos sanos, el segundo a partir de los resultados obtenidos en el primero, afinando la metodología y para un grupo de 20 niños entre 8 y 12 años, los resultados del segundo estudio sirven para justificar su aplicación en niños con TEA que presenten déficit de atención. ABSTRACT Medicine and engineering in the 21st century have resulted in advances for society but in most cases the treatments are often costly and invasive. The health education society receive is scarce, since only go to the doctor when we are really sick. With this work I present my commitment to complementary therapies, my little grain of sand in the development of a noninvasive therapeutic approach and very low cost, well and can be used in a preventive manner resulting in a society with less sick. The purpose of this thesis is to outline a methodology that can then be applied to neurological diseases, here we lay the groundwork. New researchers are needed to continue this path for a sufficiently extensive records database of subjects who have undergone binaural therapy, and so to draw firm conclusions. The contribution of this thesis includes: the application, selection, signal processing and algorithm development, indicated cognitive tests for the specific case at hand, calculation of system uncertainty of the system and development of a specific psychoacoustic test. The use of sound in medicine, such as music therapy or sound therapy has experienced a great diffusion in recent years, more than 100,000 new citations have appeared over the previous year but very few are those referring to acoustic physical characteristics of sound employee, we have only found a couple of them that physical sound characteristics are correlated with the therapeutic response. We found no specific citations posing a scientific experimental model capable of reproducing the same answers to the same parameters and stimuli. In this thesis we propose the use of binaural sound stimulation which involves the use of two identical but slightly different in frequency pure tones presented separately each in one ear, as a result the subject perceives a third tone, called binaural tone, formed by the difference in frequency with amplitude variations Studies suggest that these binaural frequencies can modify the electrical patterns of brain activity and arousal levels, being known in the literature under the name of “entrainment”. After the literature review of the state of the art, we conclude, it is necessary to develop well-designed double-blind studies, in order to establish a solid foundation on the effects of such stimulation, since most of the documented benefits relate to very small samples and unscientific may be obtained positive results due to the placebo effect. The binaural technology is cheap being any progress in this direction in the public interest. The specific objective of the research is to study the potential of binaural waves in a particular area: tasks requiring attention and concentration also we want to get any change in brain waves that can correlate with improvements. In view of the results of this research we seek to apply this methodology in neuropathology presenting any deficiency in the area of attention such as Autism Spectrum Disorder. In this thesis we present the results of two independent studies, the first to lay the foundation of the method (times, stimulation design, processing) in a sample of 78 healthy adults, the second from the results obtained in the first, refine the methodology for a group of 20 children between 8 and 12 years, the results of the second study used to justify its use in children with ASD that present attention deficit.

Development of a human movement monitoring system based on wearable devices

It is essential to remotely and continuously monitor the movements of individuals in many social areas, for example, taking care of aging people, physical therapy, athletic training etc. Many methods have been used, such as video record, motion analysis or sensor-based methods. Due to the limitations in remote communication, power consumption, portability and so on, most of them are not able to fulfill the requirements. The development of wearable technology and cloud computing provides a new efficient way to achieve this goal. This paper presents an intelligent human movement monitoring system based on a smartwatch, an Android smartphone and a distributed data management engine. This system includes advantages of wide adaptability, remote and long-term monitoring capacity, high portability and flexibility. The structure of the system and its principle are introduced. Four experiments are designed to prove the feasibility of the system. The results of the experiments demonstrate the system is able to detect different actions of individuals with adequate accuracy.

Cepstral peak prominence: a comprehensive analysis

An analytical study of cepstral peak prominence (CPP) is presented, intended to provide an insight into its meaning and relation with voice perturbation parameters. To carry out this analysis, a parametric approach is adopted in which voice production is modelled using the traditional source-filter model and the first cepstral peak is assumed to have Gaussian shape. It is concluded that the meaning of CPP is very similar to that of the first rahmonic and some insights are provided on its dependence with fundamental frequency and vocal tract resonances. It is further shown that CPP integrates measures of voice waveform and periodicity perturbations, be them either amplitude, frequency or noise.

Non uniform embedding based on relevance analysis with reduced computational complexity: application to the detection of pathologies from biosignal recordings

Nonlinear analysis tools for studying and characterizing the dynamics of physiological signals have gained popularity, mainly because tracking sudden alterations of the inherent complexity of biological processes might be an indicator of altered physiological states. Typically, in order to perform an analysis with such tools, the physiological variables that describe the biological process under study are used to reconstruct the underlying dynamics of the biological processes. For that goal, a procedure called time-delay or uniform embedding is usually employed. Nonetheless, there is evidence of its inability for dealing with non-stationary signals, as those recorded from many physiological processes. To handle with such a drawback, this paper evaluates the utility of non-conventional time series reconstruction procedures based on non uniform embedding, applying them to automatic pattern recognition tasks. The paper compares a state of the art non uniform approach with a novel scheme which fuses embedding and feature selection at once, searching for better reconstructions of the dynamics of the system. Moreover, results are also compared with two classic uniform embedding techniques. Thus, the goal is comparing uniform and non uniform reconstruction techniques, including the one proposed in this work, for pattern recognition in biomedical signal processing tasks. Once the state space is reconstructed, the scheme followed characterizes with three classic nonlinear dynamic features (Largest Lyapunov Exponent, Correlation Dimension and Recurrence Period Density Entropy), while classification is carried out by means of a simple k-nn classifier. In order to test its generalization capabilities, the approach was tested with three different physiological databases (Speech Pathologies, Epilepsy and Heart Murmurs). In terms of the accuracy obtained to automatically detect the presence of pathologies, and for the three types of biosignals analyzed, the non uniform techniques used in this work lightly outperformed the results obtained using the uniform methods, suggesting their usefulness to characterize non-stationary biomedical signals in pattern recognition applications. On the other hand, in view of the results obtained and its low computational load, the proposed technique suggests its applicability for the applications under study.

Monte Carlo limit cycle characterization

The fixed point implementation of IIR digital filters usually leads to the appearance of zero-input limit cycles, which degrade the performance of the system. In this paper, we develop an efficient Monte Carlo algorithm to detect and characterize limit cycles in fixed-point IIR digital filters. The proposed approach considers filters formulated in the state space and is valid for any fixed point representation and quantization function. Numerical simulations on several high-order filters, where an exhaustive search is unfeasible, show the effectiveness of the proposed approach.

Independent doubly Adaptive Rejection Metropolis Sampling

Adaptive Rejection Metropolis Sampling (ARMS) is a wellknown MCMC scheme for generating samples from onedimensional target distributions. ARMS is widely used within Gibbs sampling, where automatic and fast samplers are often needed to draw from univariate full-conditional densities. In this work, we propose an alternative adaptive algorithm (IA2RMS) that overcomes the main drawback of ARMS (an uncomplete adaptation of the proposal in some cases), speeding up the convergence of the chain to the target. Numerical results show that IA2RMS outperforms the standard ARMS, providing a correlation among samples close to zero.

An adaptive population importance sampler

Monte Carlo (MC) methods are widely used in signal processing, machine learning and communications for statistical inference and stochastic optimization. A well-known class of MC methods is composed of importance sampling and its adaptive extensions (e.g., population Monte Carlo). In this work, we introduce an adaptive importance sampler using a population of proposal densities. The novel algorithm provides a global estimation of the variables of interest iteratively, using all the samples generated. The cloud of proposals is adapted by learning from a subset of previously generated samples, in such a way that local features of the target density can be better taken into account compared to single global adaptation procedures. Numerical results show the advantages of the proposed sampling scheme in terms of mean absolute error and robustness to initialization.

Grouped sparsity algorithm for multichannel intracardiac ECG synchronization

In this paper, a new method is presented to ensure automatic synchronization of intracardiac ECG data, yielding a three-stage algorithm. We first compute a robust estimate of the derivative of the data to remove low-frequency perturbations. Then we provide a grouped-sparse representation of the data, by means of the Group LASSO, to ensure that all the electrical spikes are simultaneously detected. Finally, a post-processing step, based on a variance analysis, is performed to discard false alarms. Preliminary results on real data for sinus rhythm and atrial fibrillation show the potential of this approach.

Transmitter-side antennas correlation in SVD-assisted MIMO systems

MIMO techniques allow increasing wireless channel performance by decreasing the BER and increasing the channel throughput and in consequence are included in current mobile communication standards. MIMO techniques are based on benefiting the existence of multipath in wireless communications and the application of appropriate signal processing techniques. The singular value decomposition (SVD) is a popular signal processing technique which, based on the perfect channel state information (PCSI) knowledge at both the transmitter and receiver sides, removes inter-antenna interferences and improves channel performance. Nevertheless, the proximity of the multiple antennas at each front-end produces the so called antennas correlation effect due to the similarity of the various physical paths. In consequence, antennas correlation drops the MIMO channel performance. This investigation focuses on the analysis of a MIMO channel under transmitter-side antennas correlation conditions. First, antennas correlation is analyzed and characterized by the correlation coefficients. The analysis describes the relation between antennas correlation and the appearance of predominant layers which significantly affect the channel performance. Then, based on the SVD, pre- and post-processing is applied to remove inter-antenna interferences. Finally, bit- and power allocation strategies are applied to reach the best performance. The resulting BER reveals that antennas correlation effect diminishes the channel performance and that not necessarily all MIMO layers must be activated to obtain the best performance.

Resource allocation in GMD and SVD-based MIMO System

Singular-value decomposition (SVD)-based multiple-input multiple output (MIMO) systems, where the whole MIMO channel is decomposed into a number of unequally weighted single-input single-output (SISO) channels, have attracted a lot of attention in the wireless community. The unequal weighting of the SISO channels has led to intensive research on bit- and power allocation even in MIMO channel situation with poor scattering conditions identified as the antennas correlation effect. In this situation, the unequal weighting of the SISO channels becomes even much stronger. In comparison to the SVD-assisted MIMO transmission, geometric mean decomposition (GMD)-based MIMO systems are able to compensate the drawback of weighted SISO channels when using SVD, where the decomposition result is nearly independent of the antennas correlation effect. The remaining interferences after the GMD-based signal processing can be easily removed by using dirty paper precoding as demonstrated in this work. Our results show that GMD-based MIMO transmission has the potential to significantly simplify the bit and power loading processes and outperforms the SVD-based MIMO transmission as long as the same QAM-constellation size is used on all equally-weighted SISO channels.