The beekeepers need to inspect their hives regularly in order to protect them from various stressors. Manual inspection of hives require a lot of time and effort. Hence, many researchers have started using electronic beehive monitoring (EBM) systems to collect critical information from beehives, so as to alert the beekeepers of possible threats to the hive. EBM collects information by applying multiple sensors into the hive. The sensors collect information in the form of video, audio or temperature data from the hives. This thesis involves the automatic classification of audio samples from a beehive into bee buzzing, cricket chirping and ambient noise, using machine learning models. The classification of samples in these three categories will help the beekeepers to determine the health of beehives by analyzing the sound patterns in a typical audio sample from beehive. Abnormalities in the classification pattern over a period of time can notify the beekeepers about potential risk to the hives such as attack by foreign bodies (Varroa mites or wing virus), climate changes and other stressors.

Mobile devices have advanced tremendously during the last ten years and have changed our daily life in various ways. Secure pairing of mobile devices has become a significant issue considering the huge quantity of active mobile device connections and mobile traffic. However, current commonly used file sharing mobile applications rely on servers completely that are always targeted by attackers. In this thesis work, an innovative mechanism is proposed to generate symmetric keys on both mobile devices independently from a shared movement in arbitrary pattern, which means no server needs to be involved and no data exchange needed. A secret wireless-communication channel can then be established with a particular network strategy.

Techniques for improving the information quality present in imagery for feature extraction are proposed in this thesis. Specifically, two methods are presented: soft feature extraction and improved Evolution-COnstructed (iECO) features. Soft features comprise the extraction of image-space knowledge by performing a per-pixel weighting based on an importance map. Through soft features, one is able to extract features relevant to identifying a given object versus its background. Next, the iECO features framework is presented. The iECO features framework uses evolutionary computation algorithms to learn an optimal series of image transforms, specific to a given feature descriptor, to best extract discriminative information. That is, a composition of image transforms are learned from training data to present a given feature descriptor with the best opportunity to extract its information for the application at hand. The proposed techniques are applied to an automatic explosive hazard detection application and significant results are achieved.

Detecting botnets in a network is crucial because bot-activities impact numerous areas such as security, finance, health care, and law enforcement. Most existing rule and flow-based detection methods may not be capable of detecting bot-activities in an efficient manner. Hence, designing a robust botnet-detection method is of high significance. In this study, we propose a botnet-detection methodology based on graph-based features. Self-Organizing Map is applied to establish the clusters of nodes in the network based on these features. Our method is capable of isolating bots in small clusters while containing most normal nodes in the big-clusters. A filtering procedure is also developed to further enhance the algorithm efficiency by removing inactive nodes from bot detection. The methodology is verified using real-world CTU-13 and ISCX botnet datasets and benchmarked against classification-based detection methods. The results show that our proposed method can efficiently detect the bots despite their varying behaviors.

Artificial neural network (ANN) is a machine learning approach where parameters, i.e., frequency parameters and amplitude parameters, are learnt during the training process. Random features model is a special case of ANN that the structure of random features model is as same as ANN’s but the parameters’ learning processes are different. For random features model, the amplitude parameters are learnt during the training process but the frequency parameters are sampled from some distributions. If the frequency distribution of the random features model is well-chosen, both models can approximate data well. Adaptive random Fourier features with Metropolis sampling is an enhanced random Fourier features model which can select appropriate frequency distribution adaptively. This thesis studies Rectified Linear Unit and sigmoid features and combines them with the adaptive idea to generate another two adaptive random features models. The results show that using the particular set of hyper-parameters, adaptive random Rectified Linear Unit features model can also approximate the data relatively well, though the adaptive random Fourier features model performs slightly better. / I artificiella neurala nätverk (ANN), som används inom maskininlärning, behöver parametrar, kallade frekvensparametrar och amplitudparametrar, hittasgenom en så kallad träningsprocess. Random feature-modeller är ett specialfall av ANN där träningen sker på ett annat sätt. I dessa modeller tränasamplitudparametrarna medan frekvensparametrarna samplas från någon sannolikhetstäthet. Om denna sannolikhetstäthet valts med omsorg kommer båda träningsmodellerna att ge god approximation av givna data. Metoden Adaptiv random Fourier feature[1] uppdaterar frekvensfördelningen adaptivt. Denna uppsats studerar aktiveringsfunktionerna ReLU och sigmoid och kombinerar dem med den adaptiva iden i [1] för att generera två ytterligare Random feature-modeller. Resultaten visar att om samma hyperparametrar som i [1] används så kan den adaptiva ReLU features-modellen approximera data relativt väl, även om Fourier features-modellen ger något bättre resultat.

The Heyang dialect has many distinct phonological features, which make it quite different from its adjacent dialects. The phonological features of the Heyang dialect are systematically studied, and the historical strata are revealed. Diverse historical strata exist in the current system of the Heyang dialect. In the Heyang dialect, there are phonological features which belong to the stratum of the Northwestern dialect during the Tang and Song dynasties. These features include: the Middle Chinese voiced obstruents are all aspitrated; the -ŋ ending is lost in the colloquial readings of Dang (宕) and Geng (梗) rhyme groups; the division III hekou syllables in Zhi (止) and Yu (遇) rhyme groups merge; and the division III and IV hekou finals of Xie (蟹) rhyme group are xiyin. The initials yi (疑) and wei (微) in the Heyang dialect are pronounced the same as they are in the Zhongyuan yinyun. The kaikou contrasted with the hekou finals in Guo (果) rhyme group when they combined with velar and glottal initials, the division I contrasted with division II finals of Xiao (效) rhyme group in the Heyang dialect. Those phonological phenomena belong to the historical stratum of the Zhongyuan yinyun. The Heyang dialect was further compared with the Meixian dialect, a representive of the Hakka dialect group. The two dialects share so many phonological characteristics. The relation between the two dialects is even closer than that between the Heyang dialect and Mandarin, in some essential aspects, which strongly suggests that the Heyang dialect may be rooted from the Zhongyuan dialects during the Tang and Song dynasty.

The main goal of this thesis is to test the feasibility of human identification using the Electrocardiogram (ECG). Such biomedical signal has several key advantages including its intrinsic nature and liveness indicator which makes it more secure compared to some of the existing conventional and traditional biometric modalities. In compliance with the terms and regulations of McMaster University, this work has been assembled into a sandwich thesis format which consist of three journal papers. The main idea of this work is to identify individuals using distance measurement techniques and ECG feature matching. In addition, we gradually developed the content of the three papers. In the first paper, we started with the general criteria for developing ECG based biometric systems. To explain, we proposed both fiducial and non-fiducial approaches to extract the ECG features followed by providing comparative study on the performance of both approaches. Next, we applied non-overlapped data windows to extract the ECG morphological and spectral features. The former set of features include the amplitude and slope differences between the Q, R and S peaks. The later features include extracting magnitudes of the ECG frequency components using short time Fourier Transform (STFT). In addition, we proposed a methodology for QRS detection and segmentation using STFT and binary classification of ECG fiducial features. In the second paper, we proposed a technique for choosing overlapped data windows to extract the abovementioned features. Namely, the dynamic change in the ECG features from heart beats to heartbeat is utilized for identification purposes. To improve the performance of the proposed techniques we developed Frechet-mean based classifier for this application. These classifiers exploit correlation matrix structure that is not accounted for in classical Euclidean techniques. In addition to considering the center of the cluster, the Frechet-mean based techniques account for the shape of the cluster as well. In the third paper, the thesis is extended to address the variability of ECG features over multiple records. Specifically, we developed a multi-level wavelet-based filtering system which utilizes features for multiple ECGs for human identification purposes. In addition, we proposed a soft decision-making technique to combine information collected from multi-level identification channels to reach a common final class. Lastly, we evaluated the robustness of all our proposed methods over several random experiments by changing the testing data and we achieved excellent results. The results of this thesis show that the ECG is a promising biometric modality. We evaluated the performance of the proposed methods on the public ECG ID database because it was originally recorded for biometric purposes. In addition, to make performance evaluation more realistic we used two recordings of the same person obtained under possibly different conditions. Furthermore, we randomly changed both the training and testing data which are obtained from the full ECG records for performance evaluation purposes. However, it is worth mentioning that in all parts of the thesis, various parameters settings are presented to support the main ideas and it is subject to change according to human activity and application requirements. Finally, the thesis concludes with a comparison between all the proposed methods, and it provides suggestions on few open problems that can be considered for future research as extension to the work that has been done in this thesis. Generally, these problems are associated with the constraints on computational time, data volume and ECG clustering. / Thesis / Doctor of Philosophy (PhD)

Movement patterns of animals are varied, complex, and can be influenced by environmental and anthropogenic factors. One form of animal movement, migration, is influenced by environmental factors that alter the timing, duration, intensity, and likelihood of migration. Additionally, features of the landscape, both natural and anthropogenic, can alter how animals move through their seasonal and home ranges. Movement patterns can be impeded or prevented by features such as railroads, rivers, and roads. Here we explore characteristics of elk (Cervus canadensis) migration by evaluating the strategies of migration exhibited and the factors that influence migration of elk in central and eastern Utah. In addition we determine landscape features that act as barriers to movement of elk. In the first chapter, we characterize migration of elk, and examine the influence of environmental factors have on the propensity and intensity of migration. In the second chapter, we determine barriers to movement and if the identified barriers could be better used as management area boundaries.

Shape optimization has become an important tool in industry to minimize weight and generate new designs. At the same time, companies are turning to CAD-centric design strategies where robust parametric CAD models are used to generate new designs and part-families of current designs, as well as the tooling and manufacturing procedures. However, due to its complexity, the optimal topology results are often discarded or recreated by hand into a CAD model. From a design stand point, the results can be improved with the use of manufacturing constraints on the shape optimization process. These constraints improve the manufacturability based on common manufacturing practices. Even with these improvements, the process of converting topology results to CAD can cost substantial amounts of time and money. This thesis proposes a method of semi-automatically recognizing the voids, created during the shape optimization process, with parametric features based on CAD geometry construction. These parametric features are based on sets of cross-sectional shapes and spine rules to create solid objects. These features are then sent to the CAD part file via programming APIs that exist in the software packages. By recognizing features usable to the CAD systems, the voids can be characterized in the CAD model using robust dimensional constraints. This allows for the CAD model approximation to represent the topology optimization results with dimensional values from simpler shapes. Size optimization can then be applied to optimize the approximating model and regain any fidelity loss in the analytic model. Test cases created with and without manufacturing constraints show considerable promise in a proof-of-concept scenario. These tests utilize the topology optimization software HyperMesh from Altair and the CAD package NX 4.0 from Siemens (formerly UGS). The voids from shape optimization in these tests are recognized inside of HyperMesh, fit with a simple parametric feature, and created in the part model using the Open C API in NX.

Over the years, cancer has increasingly become a global health problem [12]. For successful treatment, early detection and diagnosis is critical. Radiomics is the use of CT, PET, MRI or Ultrasound imaging as input data, extracting features from image-based data, and then using machine learning for quantitative analysis and disease prediction [23, 14, 19, 1]. Feature reduction is critical as most quantitative features can have unnecessary redundant characteristics. The objective of this research is to use machine learning techniques in reducing the number of dimensions, thereby rendering the data manageable. Radiomics steps include Imaging, segmentation, feature extraction, and analysis. For this research, a large-scale CT data for Lung cancer diagnosis collected by scholars from Medical University in China is used to illustrate the dimension reduction techniques via R, SAS, and Python softwares. The proposed reduction and analysis techniques were PCA, Clustering, and Manifold-based algorithms. The results indicated the texture-based features