M/EEG mechanisms allow determining changes in the brain activity, which is useful in diagnosing brain disorders such as epilepsy. They consist of measuring the electric potential at the scalp and the magnetic field around the head. The measurements are related to the underlying brain activity by a linear model that depends on the lead-field matrix. Localizing the sources, or dipoles, of M/EEG measurements consists of inverting this linear model. However, the non-uniqueness of the solution (due to the fundamental law of physics) and the low number of dipoles make the inverse problem ill-posed. Solving such problem requires some sort of regularization to reduce the search space. The literature abounds of methods and techniques to solve this problem, especially with variational approaches. This thesis develops Bayesian methods to solve ill-posed inverse problems, with application to M/EEG. The main idea underlying this work is to constrain sources to be sparse. This hypothesis is valid in many applications such as certain types of epilepsy. We develop different hierarchical models to account for the sparsity of the sources. Theoretically, enforcing sparsity is equivalent to minimizing a cost function penalized by an l0 pseudo norm of the solution. However, since the l0 regularization leads to NP-hard problems, the l1 approximation is usually preferred. Our first contribution consists of combining the two norms in a Bayesian framework, using a Bernoulli-Laplace prior. A Markov chain Monte Carlo (MCMC) algorithm is used to estimate the parameters of the model jointly with the source location and intensity. Comparing the results, in several scenarios, with those obtained with sLoreta and the weighted l1 norm regularization shows interesting performance, at the price of a higher computational complexity. Our Bernoulli-Laplace model solves the source localization problem at one instant of time. However, it is biophysically well-known that the brain activity follows spatiotemporal patterns. Exploiting the temporal dimension is therefore interesting to further constrain the problem. Our second contribution consists of formulating a structured sparsity model to exploit this biophysical phenomenon. Precisely, a multivariate Bernoulli-Laplacian distribution is proposed as an a priori distribution for the dipole locations. A latent variable is introduced to handle the resulting complex posterior and an original Metropolis-Hastings sampling algorithm is developed. The results show that the proposed sampling technique improves significantly the convergence. A comparative analysis of the results is performed between the proposed model, an l21 mixed norm regularization and the Multiple Sparse Priors (MSP) algorithm. Various experiments are conducted with synthetic and real data. Results show that our model has several advantages including a better recovery of the dipole locations. The previous two algorithms consider a fully known leadfield matrix. However, this is seldom the case in practical applications. Instead, this matrix is the result of approximation methods that lead to significant uncertainties. Our third contribution consists of handling the uncertainty of the lead-field matrix. The proposed method consists in expressing this matrix as a function of the skull conductivity using a polynomial matrix interpolation technique. The conductivity is considered as the main source of uncertainty of the lead-field matrix. Our multivariate Bernoulli-Laplacian model is then extended to estimate the skull conductivity jointly with the brain activity. The resulting model is compared to other methods including the techniques of Vallaghé et al and Guttierez et al. Our method provides results of better quality without requiring knowledge of the active dipole positions and is not limited to a single dipole activation.

Speech communication is gaining in popularity in many different contexts as technology evolves. With the introduction of mobile electronic devices such as cell phones and laptops, and fixed electronic devices such as video and teleconferencing systems, more people are communicating which leads to an increasing demand for new services and better speech quality. Methods to enhance speech recorded by microphones often operate blindly without prior knowledge of the signals. With the addition of multiple microphones to allow for spatial filtering, many blind speech enhancement methods have to operate blindly also in the spatial domain. When attempting to improve the quality of spoken communication it is often necessary to be able to reliably determine the location of the speakers. A dedicated source localization method on top of the speech enhancement methods can assist the speech enhancement method by providing the spatial information about the sources. This thesis addresses the problem of speech-source localization, with a focus on the problem of localization in the presence of multiple concurrent speech sources. The primary work consists of methods to estimate the direction of arrival of multiple concurrent speech sources from an array of sensors and a method to correct the ambiguities when estimating the spatial locations of multiple speech sources from multiple arrays of sensors. The thesis also improves the well-known SRP-based methods with higher-order statistics, and presents an analysis of how the SRP-PHAT performs when the sensor array geometry is not fully calibrated. The thesis is concluded by two envelope-domain-based methods for tonal pattern detection and tonal disturbance detection and cancelation which can be useful to further increase the usability of the proposed localization methods. The main contribution of the thesis is a complete methodology to spatially locate multiple speech sources in enclosed environments. New methods and improvements to the combined solution are presented for the direction-of-arrival estimation, the location estimation and the location ambiguity correction, as well as a sensor array calibration sensitivity analysis.

Sound source localization is an important task for several applications and the use of deep learning for this task has recently become a popular research topic. While the majority of the previous work has focused on static sound sources, in this work we evaluate the performance of a deep learning classification system for localization of high-speed moving sound sources. In particular, we systematically evaluate the effect of a wide range of parameters at three levels including: data generation (e.g., acoustic conditions), feature extraction (e.g., STFT parameters), and model training (e.g., neural network architectures). We evaluate the performance of multiple metrics in terms of precision, recall, F-score and confusion matrix in a multi-class multi-label classification framework. We used four different deep learning models: feedforward neural networks, recurrent neural network, gated recurrent networks and temporal Convolutional neural network. We showed that (1) the presence of some reverberation in the training dataset can help in achieving better detection for the direction of arrival of acoustic sources, (2) window size does not affect the performance of static sources but highly affects the performance of moving sources, (3) sequence length has a significant effect on the performance of recurrent neural network architectures, (4) temporal convolutional neural networks can outperform both recurrent and feedforward networks for moving sound sources, (5) training and testing on white noise is easier for the network than training on speech data, and (6) increasing the number of elements in the microphone array improves the performance of the direction of arrival estimation.

No description available.

The concept of replacing threads with flexible wires and sensors in a fabric to provide an underlying platform for integrating electronic components is known as e-textiles. This concept can be used to design applications involving different types of electronic components including sensors, digital signal processors, microcontrollers, color-changing fibers, and power sources. The adaptability of the textiles to the needs of the individual and the functionality of electronics can be integrated to provide unobtrusive, robust, and inexpensive clothing with novel features. This thesis focuses on the design of e-textiles for acoustic signal processing applications. This research examines challenges encountered when developing e-textile applications involving distributed arrays of microphones. A framework for designing such applications is presented. The design process and the performance analysis of two e-textiles, a large-scale beamforming fabric and a speech-processing vest, are presented. / Master of Science

Unmanned vehicles are often used in time-critical missions such as reconnaissance or search and rescue. To this end, this thesis provides autonomous localization and mapping tools for 1/R² sources. A "1/R²" source is one in which the received intensity of the source is inversely proportional to the square of the distance from the source. An autonomous localization algorithm is developed which utilizes a particle swarm particle ltering method to recursively estimate the location of a source. To implement the localization algorithm experimentally, a command interface with Virginia Tech's autonomous helicopter was developed. The interface accepts state information from the helicopter, and returns command inputs to drive the helicopter autonomously to the source. To make the use of the system more intuitive, a graphical user interface was developed which provides localization functionality as well as a waypoint navigation outer-loop controller for the helicopter. This assists in positioning the helicopter and returning it home after the the algorithm is completed. An autonomous mapping mission with a radioactive source is presented, along with a localization experiment utilizing simulated sensor readings. This work is the rst phase of an on-going project at the Unmanned Systems Lab. Accordingly, this thesis also provides a framework for its continuation in the next phase of the project. / Master of Science

A rise in interest for service robotic rovers produces a need for a low-cost method for source localization in order for a prospective robotic unit to engage with a human operator. This study examines the use of the LMS algorithm for constructing a beamformer using an optimized Weiner filter solution for this source localization application and evaluates the robustness of a developed characterization method for assuring that a proper approximation for the desired signal is achieved. The method presented in this paper encompasses using a filter and sum method in which the sums are generated for a selected set of filter angles, and this set of sums are compared and characterized to produce a selection for an approximate arrival angle from the sound source to the microphone array. These filters are adaptively trained offline using a generated desired signal chirp to represent the average human whistle and a training data set for each of the four possible room configurations. This method was tested to determine if a selected filter configuration could still produce viable outputs for scenarios in which the testing room had been changed, whether noise was injected into the testing environment, if two or three microphones were used in testing process, and whether the filter angles are aligned with the arrival angles of the signal. Results on the robustness of the adaptive LMS beamformer are presented. Limitations of the system performance are discussed and possible solutions for results that have undesired performance are given in future work. / Master of Science / A rise in interest for service robotic rovers produces a need for a low-cost method for locating a sound source so that a potential service robot can interact with a human operator. In this study, a beamformer is implemented to approximate a direction for the sound source. This beamformer is comprised of a set of trained filters for the designed microphone array. These filters were trained based on three training conditions of training room, the number of microphones used, and whether additive or ambient noise is used during training. The training signal for the filters consisted of a chirp from 1 to 2.5 kHz to mimic a portion of the human whistling spectrum. Once trained, these beamformers were then given data from separate tests to determine if a distinct and correct approximation could be determined. This paper suggests a method to use the correlation of each beamformer to the training signal to determine both the maximum correlated beamformer and whether correlation is distinct from greater than the other beamformers examined. These results are finally examined under an ANOVA and percent difference process to determine if the three training conditions improve the average prediction of the angle of arrival of the source signal for the generated beamformers.

Detection and localization of sound sources in an ocean environment can be achieved with a distributed array of passive acoustic sensors. Utilizing an array of autonomous littoral gliders, which offer long-term and quiet operation, and vector sensors, which measure both acoustic pressure and particle velocity, the array's localization performance can be improved. However, vector sensors are susceptible to errors induced by acoustic noise, and autonomous gliders as a sensor platform introduce positional errors. Through both simulations and at-sea data, the localization performance of four processing methods are evaluated under various noisy conditions. In both simulated and at-sea data results, a new cross-coherent method outperforms traditional methods by mitigating the effects of acoustic noise, provided sufficient positional accuracy of the array elements.

Acoustic source localization (ASL) is pinpointing an acoustic source. ASL can reveal the point of impact of a foreign object or the point of crack initiation in a structure. ASL is necessary for continuous health monitoring of a structure. ASL in an anisotropic plate is a challenging task. This dissertation aims to investigate techniques that are currently being used to precisely determine an acoustic source location in an anisotropic plate without knowing its material properties. A new technique is developed and presented here to overcome the existing shortcomings of the acoustic source localization in anisotropic plates. It is done by changing the analysis perspective from the angular dependent group velocity of the wave and its straight line propagation to the wave front shapes and their geometric properties when a non-circular wave front is generated. Especially, 'rhombic wave front' and 'elliptical wave front' are dealt with because they are readily observed in highly anisotropic composite plates. Once each proposed technique meets the requirements of measurement, four sensor clusters in three different quadrants (recorded by 12 sensors) for the rhombus and at least three sensor clusters (recorded by 9 sensors) for the ellipse, accurate Acoustic Source Localization is obtained. It has been successfully demonstrated in the numerical simulations. In addition, a series of experimental tests demonstrate reliable and robust prediction performance of the developed new acoustic source localization technique.

ITC/USA 2013 Conference Proceedings / The Forty-Ninth Annual International Telemetering Conference and Technical Exhibition / October 21-24, 2013 / Bally's Hotel & Convention Center, Las Vegas, NV / Relay selection for optimal communication as well as multiple source localization is studied. We consider the use of dual-role nodes that can work both as relays and also as anchors. The dual-role nodes and multiple sources are placed at fixed locations in a two-dimensional space. Each dual-role node estimates its distance to all the sources within its radius of action. Dual-role selection is then obtained considering all the measured distances and the total SNR of all sources-to-destination channels for optimal communication and multiple source localization. Bit error rate performance as well as mean squared error of the proposed optimal dual-role node selection scheme are presented.