京都大学 / 0048 / 新制・課程博士 / 博士(医科学) / 甲第18905号 / 医科博第61号 / 新制||医科||4(附属図書館) / 31856 / 京都大学大学院医学研究科医科学専攻 / (主査)教授 河本 宏, 教授 武田 俊一, 教授 髙折 晃史 / 学位規則第4条第1項該当 / Doctor of Medical Science / Kyoto University / DFAM

Signal detection theory (SDT) has proven to be a robust and useful statistical model for analyzing human performance in detection and decision making tasks. As with many models extensions have been proposed in order capture and represent the real world to a greater degree. Multidimensional Signal Detection Theory (MSDT) has had success in describing and modeling complex signals, signals that are comprised by more than one identifiable component dimension. Fuzzy Signal Detection Theory (FSDT) has had success in modeling and measuring human performance in cases where there exist ambiguity in the signal or response dimension characteristics, through the application of fuzzy set theory to the definition of the performance outcome categories. Multidimensional Fuzzy Signal Detection Theory (MFSDT) was developed to accommodate simultaneously both the multidimensionality of a signal and the fuzzification of outcome categories in order to integrate the two extensions. A series of three studies were performed to develop and test the theory. One study's purpose was to develop and derive multidimensional mapping functions, the aspect of MFSDT where MSDT and FSDT were integrated. Two receiver operating characteristic (ROC) studies were performed, one simulated and one empirical. The results from both ROC analysis indicated that for perceptually separable and perceptually integral complex stimuli that MFDST is a viable methodological approach to analyzing performance of signal detection tasks where there are complex signals with ambiguous signal characteristics.

This thesis develops a model of the Cal Poly Wind Turbine that is used to determine if there is an imbalance in the turbine rotor. A theoretical model is derived to estimate the expected vibrations when there is an imbalance in the rotor. Vibration and acceleration data are collected from the turbine tower during operation to confirm the model is useful and accurate for determining imbalances in the turbine. Digital signal processing techniques for analyzing the vibration data are explored and tested with simulation data. This includes frequency shifts, lock-in amplifiers, phase-locked loops, discrete Fourier transforms, and decimation filters. The processed data is fed into an algorithm that determines if there is an imbalance. The detection algorithm consists of a machine learning classification model that uses experimental data to train and increase the success rate of the imbalance detection. Various models are explored, including the K-Nearest Neighbors algorithm, logistic regression, and neural networks. These models have trade-offs between mathematical complexity, required computing power, scalability, and accuracy. With proper implementations of these detection models, the imbalance detection accuracy was measured to be about 90%.

Graph Signal Processing (GSP) has been used in the analysis of functional Magnetic Resonance Imaging(fMRI). As a holistic view of brain function and the connections between and within brain regions, by structuring data as node points within the brain and modelling the edge connections between nodes. Many studies have used GSP with Blood Oxygenation Level Dependent (BOLD) imaging of the brain and brain activation. Meanwhile, the methodology has seen little use in muscle imaging. Similar to brain BOLD, muscle BOLD (mBOLD) also aims to demonstrate muscle activation. Muscle BOLD depends on oxygenation, vascularization, fibre type, blood flow, and haemoglobin count. Nevertheless the mBOLD signal still follows muscle activation closely. Electromyography (EMG) is another modality for measuring muscle activation. Both mBOLD and EMG can be represented and analyzed with GSP. In order to better understand muscle activation during contraction the proposed method focused on using GSP to model mBOLD data both alone and jointly with EMG. Simultaneous mBOLD imaging and EMG recording of the calf muscles was performed, creating a multimodal dataset. A generalized filtering methodology was developed for the removal of the MRI gradient artifact in EMG sensors within the MR bore. The filtered data was then used to generate a GSP model of the muscle, focusing on gastrocnemius, soleus, and tibialis anterior muscles. The graph signals were constructed along two edge connection dimensions; coherence and fractility. For the standalone mBOLD graph signal models, the models’ goodness of fits were 1.3245 × 10-05 and 0.06466 for coherence and fractility respectively. The multimodal models showed values of 2.3109 × -06 and 0.0014799. These results demonstrate the promise of modelling muscle activation with GSP and its ability to incorporate multimodal data into a singular model. These results set the stage for future investigations into using GSP to represent muscle with mBOLD, EMG, and other biosignal modalities. / Thesis / Master of Applied Science (MASc) / Magnetic Resonance Imaging(MRI) and electromyography (EMG) are techniques used in the analysis of muscle, for detecting injury or deepening the understanding of muscle function. Graph Signal Processing (GSP) is a methodology used to represent data and the information flow between positions. While GSP has been used in modelling the brain, applications to muscle are scarce. This work aimed to model muscle activation using GSP methods, using both MRI and EMG data. To do so, a method for being able to simultaneously record MRI and EMG data was developed through hardware construction and the software implementation of EMG signal filtering. The collected data were then used to construct multiple GSP models based on the coherence and complexity of the signals, the goodness of fit for each of the constructed models were then compared. In conclusion, it is feasible to use GSP to model muscle activity with multimodal MRI and EMG data. This shows promise for future investigations into the applications of GSP to muscle research.

Performing non-destructive testing (NDT) on materials is a helpful tool for maintenance and quality control because the materials are not destroyed or disturbed; ultrasound imaging is one type of NDT. Ultrasound imaging of heterogeneous materials contains many echoes from the material itself. These echoes come from changes in the acoustic impedance, i.e. changes in the relation between the density and the sound speed of the material. However, these echoes will show speckle characteristics in images, making it hard to detect any defects in the imaged material. In this work, a method of suppressing this speckle noise is proposed. The proposed method is a 2D Wiener filter, which with the help of an image of the healthy material models changes in the material when a new image is taken later. The filter models the changes of the speckle noise between images of a defected- and healty material and then supresses the speckle from the image with defects. The filter works well on the artificial images used in this work but have yet to be tested on actual data. A version of a weighted moving average filter was also looked into, but this filter did not produce usable results.

No description available.

No description available.

<p>We examine the robust downlink beamforming design from the point of outage probability constraint. We further reason that since the estimated downlink channel correlation (DCC) matrices form a manifold in the signal space, the estimation error should be measured in terms of Riemannian distance (RD) instead of the commonly used Euclidean distance (ED). Applying this concept of measure to our design constraint, we establish approximated outage probability constraints using multidimensional ball set and multidimensional cube set. We transform the design problem into a convex optimization problem which can be solved efficiently by standard methods. Our proposed methods apply to both Gaussian distribution assumption and uniform distribution assumption. Simulation results show that the performance of our design is superior to those of other robust beamformers recently developed.</p> / Master of Applied Science (MASc)

<p>We consider the problem of power allocation for the single-cell multiple-input single- output (MISO) downlink in a time division duplex (TDD) system. In such systems, the base station (BS) acquires information about the channel state during the training component of the uplink phase. The resulting estimation errors are modeled prob- abilistically, and the receivers specify quality-of-service (QoS) constraints in terms of a target signal-to-interference-and-noise ratio that is to be achieved with a given outage probability. For a fixed beamforming structure, we seek a power allocation that minimizes the transmission power required to satisfy the users’ QoS requests.</p> <p>The proposed approach to that problem begins with the observation that for TDD systems the channel estimation error at the base station can be modeled as being additive and Gaussian. Under that model, we obtain a precise deterministic characterization of the outage probability, and mildly conservative approximations thereof. Although the resulting deterministic optimization problems are not convex, we have been able to obtain good solutions using straightforward coordinate update algorithms. In fact, these solutions provide significantly better performance than the existing approaches, which are based on convex restrictions, because the proposed approximations are less conservative. By developing some approximations of the precise deterministic characterization of the outage probability, we develop algorithms that have good performance and much lower computational cost.</p> / Master of Applied Science (MASc)

Microphone arrays have seen wide usage in a variety of fields; especially in sonar, acoustic source monitoring and localization, telecommunications, and diagnostic medicine. The goal of most of these applications is to detect or extract a signal of interest. This task is complicated by the presence of interferers and noise, which corrupts the recorded array signals. This dissertation explores two new techniques that increase signal clarity: interferer suppression and emergent signal detection. Spatial processing is often used to suppress interferers that are spatially distinct from the signal of interest. If the signal of interest and the interferer are statistically independent, blind source separation can be used to statistically extract the signal of interest. The first new method to improve signal clarity presented in this work combines spatial processing with blind source separation to suppress interferers. This technique allows for the separation of independent sources that are not necessarily simultaneously mixed or spatially distinct. Simulations and experiments are used to show the capability of the new algorithm for a variety of conditions. The major contributions in this dissertation under this topic are to use independent component analysis to extract the signal of interest from a set of array signals, and to improve existing independent component analysis algorithms to allow for time delayed mixing. This dissertation presents a novel method of improving signal clarity through emergent signal detection. By determining which time frames contain the signal of interest, frames that contain only interferers and noise can be eliminated. When a new signal of interest emerges in a measurement of a mixed set of sources, the principal component subspace is altered. By examining the change in the subspace, the emergent signal can be robustly detected. This technique is highly effective for signals that have a near constant sample variance, but is successful at detecting a wide variety of signals, including voice signals. To improve performance, the algorithm uses a feed-forward processing technique. This is helpful for the VAD application because voice does not have a constant sample variance. Experiments and simulations are used to demonstrate the performance of the new technique. / Ph. D.