229 research outputs found
Some Estimations of Kraft Numbers and Related Results
No full textSome inequalities for Kraft numbers which are important in coding theory, for they lead to a simple criterion to determine whether or not there is an instantaneous code with given codeword lengths, are pointed out
Automatic eyeblink and muscular artifact detection and removal from EEG signals using k-nearest neighbor classifier and long hhort-term memory networks
No full textElectroencephalogram (EEG) is often corrupted with artifacts originating from sources such as eyes and muscles. Hybrid artifact removal methods often require human intervention for the adjustment of different parameters. We propose a robust method that can automatically detect and remove eyeblink and muscular artifacts from EEG using a k-nearest neighbor (kNN) classifier and a long short-term memory (LSTM) network. Our method adopts a sliding window of 0.5 s to detect and remove the artifacts from EEG. Features, such as the variance, peak-to-peak amplitude, and average rectified value, are calculated for each EEG segment to identify corrupted segments using the kNN classifier. The kNN classifier detects the presence of artifacts, after which the corresponding EEG window is forwarded to the LSTM network for artifact removal. The LSTM network is trained with the corrupted segments of 0.5 s as input and clean segments of 0.5 s as output. Our method achieved an accuracy of 97.4% in identifying corrupted EEG segments and an average correlation coefficient, structural similarity, signal-to-artifact ratio, and normalized mean squared error of 0.69, 0.76, 1.52 dB, and 0.0013, respectively, in cleaning the EEG. Our results outperformed other hybrid methods reported in the literature based on a combination of ensemble empirical mode decomposition and canonical correlation analysis, a combination of independent component analysis and wavelet decomposition, and tensor decomposition. The mean absolute error of our method is also better in comparison to other methods. Our method can be applied to single and multiple channels and does not require any tuning of parameters
Recommended from our members
Garo Bhasha
No full textGrammar of the Garo language written in Bengali including word lists and example sentences intended to help non-Garos learn the language
The impact of social media engagement on university student recruitment
No full textEducation marketing managers are increasingly becoming aware of the benefits of social media engagement in recruitment marketing for colleges and universities. Today’s prospective student is tech-savvy and information driven. Despite a general emphasis on social media engagement in a university marketing strategy, there is minimal research on the influence it has on university student recruitment and the amount of effort that universities should dedicate to social media engagement. Therefore, this study focuses on the influence of university social media engagement, particularly Facebook, on university student recruitment in the form of student applications.education marketingsocial mediarecruitmentfacebooklinear regression method
Aseismic Design of Structure-Equipment Systems Using Variable Frequency Pendulum Isolator
No full textThe assessment of association between insomnia and risk factors in the province of Nova Scotia.
No full textScientific research on sleep problems reveals that insomnia has considerable impact on the daily functioning of the affected individual. This thesis investigates risk factors associated with insomnia in a random sample of individuals aged 12 years and older (n = 5,018) living in the province of Nova Scotia, Canada. A binomial logistic regression was used to assess the association between insomnia and health factors, demographic and socioeconomic characteristics, and lifestyle variables. Moreover, the bootstrap method was used to obtain estimates of the odds ratios, parameter estimates, and their standard errors in addition to the logistic regression. The following findings are reported. Age less than 65 years, female gender, current level of smoking, arthritis, asthma, back problems, high blood pressure, bowel syndrome, diabetes, heart disease and migraine were associated with an increased risk of insomnia in the population of Nova Scotia. The bootstrap estimates for odds ratios were slightly higher than those obtained by the classical logistic regression model. This is the first study conducted for insomnia in Nova Scotia for the years 2000-2010. --P. i.The original print copy of this thesis may be available here: http://wizard.unbc.ca/record=b173787
Personalized fall detection monitoring system based on user movements
No full textDissertation (MEng)--University of Pretoria, 2018.High accuracy, fall detection systems is a fundamental requirement among the increasing elderly
population, mainly due to expensive healthcare and a shortage of nurses for home-care. Fall detection
systems have evolved over the past few years, from a button pendant to three newer types of fall
detection systems - wearable sensors, ambient sensors, and camera-based sensors. Wearable sensors
are regarded as the most popular, as it provides both indoor and outdoor monitoring and is the least
expensive among the newer fall detection systems. Detection of a fall, using wearable sensors, started
off at first by using a threshold method, where the features extracted from the wearable sensor data
are compared to a pre-defined value. The problem with this approach is that the pre-defined value
only works on a small set of people with certain user characteristics. It was also difficult to set a value
that can distinguish between everyday activities and fall activities. To solve this problem, supervised
machine learning algorithms were incorporated - these obtained higher accuracies when compared
to the threshold method. Supervised machine learning algorithms achieved high accuracy during
laboratory experiments. In a practical scenarios, the performance of these fall detections were low, due
to the supervised machine learning algorithms making use of simulated fall data which is performed
on a soft mattress which does not represent a real fall event (which is usually spontaneous). Since it is
difficult to obtain real fall data, a lot of studies make use of simulation data. Using artificial fall as
training data can result in over-fitting, which causes poor decisions. Both threshold and supervised
classifiers cannot provide a user-specific solution for each individual user. Since supervised machine
learning algorithms require everyday activities and fall activities to classify, as well as limited fall data
(which creates an imbalance i.t.o classification), it is hard for these algorithms to classify accurately.
Another problem is that these systems are limited to a certain number of activities that a user can
perform, and it does not work for everybody. User-specific personalization can be provided using
unsupervised machine learning algorithms, resulting in the following advantages: a) more activities
can be included in the classifier, and b) the fall detection system can address the inter-individual
differences. In this research, the effects of personalization models using user movements are analysed
(in terms of accuracy). A low-cost smartphone accelerometer sensor was used in the system. The study
was divided into two parts: a simulation phase and an experimental phase. The simulation phase made
use of a public dataset known as the tFall dataset. The type of input data to be used, which machine
learning algorithm to use and the different types of personalization models, were investigated. For
the type of input data, the following were considered: raw accelerometer values, statistical features
extracted from the accelerometer, principal component analysis on the statistical features extracted, or
the statistical features selected from the variance-threshold feature selection method. Both supervised
and unsupervised machine learning algorithms were implemented to determine the best algorithm.
The following unsupervised machine learning algorithms were implemented: nearest neighbour, oneclass
support vector machine, angle based outlier detection, and isolation forest. Angle based outlier
detection and isolation forest were not implemented in any fall detection systems before. For the
supervised machine learning algorithm, the two most popular machine learning algorithms were
selected: support vector machine learning algorithm and k-NN. The following models were tested: a)
Model 1 the classifier itself; b) Model 2 the non-fall activity is retrained whenever the classifier
correctly detects a non-fall activity; c) Model 3 the false positive is retrained when the classifier detects
a non-fall activity as a fall activity, and d) Model 4 combining Model 2 and Model 3. The unsupervised
machine learning algorithm is applied to all the models, whereas supervised machine learning algorithm
is only applied to Model 1. During the simulation phase, the following evaluation parameters were
used: sensitivity, specificity, geometric mean and F1-measure. During the experimental phase, the best
input data set (raw accelerometer values), model (Model 4) and classifier (angle based outlier detection)
were implemented on an Android smartphone, to demonstrate how accurately the fall detection can
classify. From experimental results, it was shown that personalization models using user movements
can improve the overall performance of the system, achieving a sensitivity of 90.48% and specificity of 92.31%.Electrical, Electronic and Computer EngineeringMEngUnrestricte
Copula based rigid-body image registration.
No full textThe literature presents a wide number of algorithms in the field of image registration. However, analysis of the literature revealed that much emphasis has not been placed on copula based image registration. Thus this thesis seeks to explain the image registration problem and how it may be solved using copula based measures. Here we are aiming to combine the MATLAB fminsearch optimization method with copula based alignment measure, in order to monitor the performance of copula based alignment measure in image registration. Performance of four copula functions namely, Clayton, Frank, Gaussian and Marshal-Olkin are tested in image registration algorithm. A comparison is then posited of the performance of the four copula functions in image registration. These four copula measures are then compared with the well known method of image registration alignment measure, that is the joint histogram based mutual information. The accuracy and speed of the image registration algorithm was monitored on aerial and medical (MRI) images. Since we are using rigid-body transformations, the image registration algorithm is categorized as rigid body image registration. --Leaf i.The original print copy of this thesis may be available here: http://wizard.unbc.ca/record=b184486
Oxidation of tetracaine hydrochloride by chloramine-b in acid medium: Kinetic modeling
No full textTetracaine hydrochloride (TCH) is one of the potent local anaesthetics. A kinetic study of oxidation of tetracaine hydrochloride by sodium N-chlorobenzenesulfonamide (chloramine-B or CAB) has been carried in HClO 4 medium at 303 K. The rate shows first-order dependence on CAB o, shows fractional-order dependence on substrate o, and is self-governing on acid concentration. Decrease of dielectric constant of the medium, by adding methanol, increased the rate. Variation of ionic strength and addition of benzenesulfonamide or NaCl have no significant effect on the rate. The reaction was studied at different temperatures and the activation parameters have been evaluated. The stoichiometry of the reaction was found to be 1: 5 and the oxidation products were identified by spectral analysis. The conjugate free acid C6H5SO2NHCl of CAB is postulated as the reactive oxidizing species. The observed results have been explained by plausible mechanism and the related rate law has been deduced. © 2014 Jayachamarajapura Pranesh Shubha and Puttaswamy
- …
