DeFFusion: CNN-based Continuous Authentication Using Deep Feature Fusion

Smartphones have become crucial and important in our daily life, but the security and privacy issues have been major concerns of smartphone users. In this article, we present DeFFusion, a CNN-based continuous authentication system using Deep Feature Fusion for smartphone users by leveraging the accelerometer and gyroscope ubiquitously built into smartphones. With the collected data, DeFFusion first converts the time domain data into frequency domain data using the fast Fourier transform and then inputs both of them into a designed CNN, respectively. With the CNN-extracted features, DeFFusion conducts the feature selection utilizing factor analysis and exploits balanced feature concatenation to fuse these deep features. Based on the one-class SVM classifier, DeFFusion authenticates current users as a legitimate user or an impostor. We evaluate the authentication performance of DeFFusion in terms of impact of training data size and time window size, accuracy comparison on different features over different classifiers and on different classifiers with the same CNN-extracted features, accuracy on unseen users, time efficiency, and comparison with representative authentication methods. The experimental results demonstrate that DeFFusion has the best accuracy by achieving the mean equal error rate of 1.00% in a 5-second time window size.

2D Wavelet Tree Ordering Based Localized Total Variation Model for Efficient Image Restoration

The degraded image during the process of image analysis needs more number of iterations to restore it. These iterations take long waiting time and slow scanning, resulting in inefficient image restoration. A few numbers of measurements are enough to recuperate an image with good condition. Due to tree sparsity, a 2D wavelet tree reduces the number of coefficients and iterations to restore the degraded image. All the wavelet coefficients are extracted with overlaps as low and high sub-band space and ordered them such that they are decomposed in the tree ordering structured path. Some articles have addressed the problems with tree sparsity and total variation (TV), but few authors endorsed the benefits of tree sparsity. In this paper, a spatial variation regularization algorithm based on tree order is implemented to change the window size and variation estimators to reduce the loss of image information and to solve the problem of image smoothing operation. The acceptance rate of the tree-structured path relies on local variation estimators to regularize the performance parameters and update them to restore the image. For this, the Localized Total Variation (LTV) method is proposed and implemented on a 2D wavelet tree ordering structured path based on the proposed image smooth adjustment scheme. In the end, a reliable reordering algorithm proposed to reorder the set of pixels and to increase the reliability of the restored image. Simulation results clearly show that the proposed method improved the performance compared to existing methods of image restoration.

When Elon Musk Changes his Tone, Does Bitcoin Adjust Its Tune?

We present a textual analysis that explains how Elon Musk’s sentiments in his Twitter content correlates with price and volatility in the Bitcoin market using the dynamic conditional correlation-generalized autoregressive conditional heteroscedasticity model, allowing less sensitive to window size than traditional models. After examining 10,850 tweets containing 157,378 words posted from December 2017 to May 2021 and rigorously controlling other determinants, we found that the tone of the world’s wealthiest person can drive the Bitcoin market, having a Granger causal relation with returns. In addition, Musk is likely to use positive words in his tweets, and reversal effects exist in the relationship between Bitcoin prices and the optimism presented by Tesla’s CEO. However, we did not find evidence to support linkage between Musk’s sentiments and Bitcoin volatility. Our results are also robust when using a different cryptocurrency, i.e., Ether this paper extends the existing literature about the mechanisms of social media content generated by influential accounts on the Bitcoin market.

Biospeckle Activity of Highbush Blueberry Fruits Infested by Spotted Wing Drosophila (Drosophila suzukii Matsumura)

In this study, the potential of the biospeckle phenomenon for detecting fruit infestation by Drosophila suzukii was examined. We tested both graphical and analytical approaches to evaluate biospeckle activity of healthy and infested fruits. As a result of testing the qualitative approach, a generalized difference method proved to be better at identifying infested areas than Fujii’s method. Biospeckle activity of healthy fruits was low and increased with infestation development. It was found that the biospeckle activity index calculated from spatial-temporal speckle correlation of THSP was the best discriminant of healthy fruits and fruits in two different stages of infestation development irrespective of window size and pixel selection strategy adopted to create the THSP. Other numerical indicators of biospeckle activity (inertia moment, absolute value of differences, average differences) distinguished only fruits in later stage of infestation. Regular values of differences turned out to be of no use in detecting infested fruits. We found that to provide a good representation of activity it was necessary to use a strategy aimed at random selection of pixels gathered around the global maximum of biospeckle activity localized on the graphical outcome. The potential of biospeckle analysis for identification of highbush blueberry fruits infested by D. suzukii was confirmed.

AI-Based Quantification of Fitness Activities Using Smartphones

To encourage more active activities that have the potential to significantly reduce the risk of people’s health, we aim to develop an AI-based mobile app to identify four gym activities accurately: ascending, cycling, elliptical, and running. To save computational cost, the present study deals with the dilemma of the performance provided by only a phone-based accelerometer since a wide range of activity recognition projects used more than one sensor. To attain this goal, we derived 1200 min of on-body data from 10 subjects using their phone-based accelerometers. Subsequently, three subtasks have been performed to optimize the performances of the K-nearest neighbors (KNN), Support Vector Machine (SVM), Shallow Neural Network (SNN), and Deep Neural Network (DNN): (1) During the process of the raw data converted to a 38-handcrafted feature dataset, different window sizes are used, and a comparative analysis is conducted to identify the optimal one; (2) principal component analysis (PCA) is adopted to extract the most dominant information from the 38-feature dataset described to a simpler and smaller size representation providing the benefit of ease of interpreting leading to high accuracy for the models; (3) with the optimal window size and the transformed dataset, the hyper-parameters of each model are tuned to optimal inferring that DNN outperforms the rest three with a testing accuracy of 0.974. This development can be further implemented in Apps Store to enhance public usage so that active physical human activities can be promoted to enhance good health and wellbeing in accordance with United Nation’s sustainable development goals.

XHAC

Explainable artificial intelligence (XAI) is a concept that has emerged and become popular in recent years. Even interpretation in machine learning models has been drawing attention. Human activity classification (HAC) systems still lack interpretable approaches. In this study, an approach, called eXplainable HAC (XHAC), was proposed in which the data exploration, model structure explanation, and prediction explanation of the ML classifiers for HAR were examined to improve the explainability of the HAR models' components such as sensor types and their locations. For this purpose, various internet of things (IoT) sensors were considered individually, including accelerometer, gyroscope, and magnetometer. The location of these sensors (i.e., ankle, arm, and chest) was also taken into account. The important features were explored. In addition, the effect of the window size on the classification performance was investigated. According to the obtained results, the proposed approach makes the HAC processes more explainable compared to the black-box ML techniques.

Environment-Specific Genomic Prediction Ability in Maize Using Environmental Covariates Depends on Environmental Similarity To Training Data

Technology advances have made possible the collection of a wealth of genomic, environmental, and phenotypic data for use in plant breeding. Incorporation of environmental data into environment-specific genomic prediction (GP) is hindered in part because of inherently high data dimensionality. Computationally efficient approaches to combining genomic and environmental information may facilitate extension of GP models to new environments and germplasm, and better understanding of genotype-by-environment (G × E) interactions. Using genomic, yield trial, and environmental data on 1,918 unique hybrids evaluated in 59 environments from the maize Genomes to Fields project, we determined that a set of 10,153 SNP dominance coefficients and a 5-day temporal window size for summarizing environmental variables were optimal for GP using only genetic and environmental main effects. Adding marker-by-environment variable interactions required dimension reduction, and we found that reducing dimensionality of the genetic data while keeping the full set of environmental covariates was best for environment-specific GP of grain yield, leading to an increase in prediction ability of 2.7% to achieve a prediction ability of 80% across environments when data were masked at random. We then measured how prediction ability within environments was affected under stratified training-testing sets to approximate scenarios commonly encountered by plant breeders, finding that incorporation of marker-by-environment effects improved prediction ability in cases where training and test sets shared environments, but did not improve prediction in new untested environments. The environmental similarity between training and testing sets had a greater impact on the efficacy of prediction than genetic similarity between training and test sets.

Whole Exome-Sequencing of Pooled Genomic DNA Samples to Detect Quantitative Trait Loci in Esotropia and Exotropia of Strabismus in Japanese

Background: Esotropia and exotropia are two major phenotypes of comitant strabismus. It remains controversial whether esotropia and exotropia would share common genetic backgrounds. In this study, we used a quantitative trait locus (QTL)-sequencing pipeline for diploid plants to screen for susceptibility loci of strabismus in whole exome sequencing of pooled genomic DNAs of individuals. Methods: Pooled genomic DNA (2.5 ng each) of 20 individuals in three groups, Japanese patients with esotropia and exotropia, and normal members in the families, was sequenced twice after exome capture, and the first and second sets of data in each group were combined to increase the read depth. The SNP index, as the ratio of variant genotype reads to all reads, and Δ(SNP index) values, as the difference of SNP index between two groups, were calculated by sliding window analysis with a 4 Mb window size and 10 kb slide size. The rows of 200 “N”s were inserted as a putative 200-b spacer between every adjoining locus to depict Δ(SNP index) plots on each chromosome. SNP positions with depth <20 as well as SNP positions with SNP index of <0.3 were excluded. Results: After the exclusion of SNPs, 12,242 SNPs in esotropia/normal group and 12,108 SNPs in exotropia/normal group remained. The patterns of the Δ(SNP index) plots on each chromosome appeared different between esotropia/normal group and exotropia/normal group. When the consecutive groups of SNPs on each chromosome were set at three patterns: SNPs in each cytogenetic band, 50 consecutive sliding SNPs, and SNPs in 4 Mb window size with 10 kb slide size, p values (Wilcoxon signed rank test) and Q values (false discovery rate) in a few loci as Manhattan plots showed significant differences in comparison between the Δ(SNP index) in the esotropia/normal group and exotropia/normal group. Conclusions: The pooled DNA sequencing and QTL mapping approach for plants could provide overview of genetic background on each chromosome and would suggest different genetic backgrounds for two major phenotypes of comitant strabismus, esotropia and exotropia.

Noise Reduction, Enhancement and Classification for Sonar Images

Ultrasound imaging has some problems with image properties output. These affects the specialist decision. Ultrasound noise type is the speckle noise which has a grainy pattern depending on the signal. There are two parts of this study. The first part is the enhancing of images with adaptive Weiner, Lee, Gamma and Frost filters with 3x3, 5x5, and 7x7 sliding windows. The evaluated process was achieved using signal to noise ratio (SNR), peak signal to noise ratio (PSNR), mean square error (MSE), and maximum difference (MD) criteria. The second part consists of simulating noise in a standard image (Lina image) by adding different percentage of speckle noise from 0.01 to 0.06. The supervised classification based minimum distance method is used to evaluate the results depending on selecting four blocks located at different places on the image. Speckle noise was added with different percentage from 0.01 to 0.06 to calculate the coherent noise within the image. The coherent noise was concluded from the slope of the standard deviation with the mean for each noise. The results showed that the additive noise increased with the slide window size, while multiplicative noise did not change with the sliding window nor with increasing noise ratio. Wiener filter has the best results in enhancing the noise.

A sliding-window based signal processing method for characterizing clusters in gas-solids high-density CFB reactor

Particle clusters in CFB risers were identified from the instantaneous solids holdup signals by a new sliding-window based signal processing method. By shifting the sliding time window and calculating the mean and the standard deviation within it, a non-linear threshold curve for identifying the clusters was derived instead of the conventional constant threshold. The optimal sliding window size was determined as Wb = 1024 data points based on the bisection method on the entire piece of signals. Using the proposed method, a more realistic characterization of the clusters in both the HDCFB and LDCFB was obtained by considering the bulk fluctuation of the gas-solids flow. The clusters in the HDCFB have higher solids holdup and lower velocity than that in the LDCFB. The HDCFB is also found to have a greater number of loose clusters for better gas-solids contacting and exchanges in the center of the riser.