This comprehensive review paper seeks to provide an in-depth survey of the technologies and methodologies employed in decision support systems for ballistocardiography. The paper extensively covers the biometric information embedded in the measured ballistocardiography signals. The presented exploration of ballistocardiography covers various pivotal stages, including signal measurement techniques, pre-processing methodologies, feature extraction approaches, classification techniques, and evaluation methods. Within the scope of this study, a systematic review has been performed, bringing together notable strategies employed in ballistocardiography from its initial stages to its current state of advancement. The efficacy of these systems in estimating ballistocardiography-based biometrics has demonstrated noteworthy proximity to acceptable levels. The utilization of ballistocardiography signals holds significant promise as an evolving field of research. This paper concludes by addressing the limitations inherent in the current state of research, outlining the potential directions for future investigations and real-world applications, and discussing the crucial aspect of explainability, which represents one of the new trends in computer-aided detection requirements.

One-class classification (OCC) is a supervised classification problem where the training data is solely one class. OCC cannot execute hyperparameter tuning because its evaluation requires access to other classes; the algorithm will no longer be OCC if the model is updated after accessing other classes. To address this issue, this paper proposes hyperparameter fusion, which is applicable without the evaluation. The fusion process applies ensemble learning techniques, voting, and stacking into OCC models trained on different hyperparameters. The experiments involve 54 OCC problems from 27 imbalanced learn datasets and 115 hyperparameter candidates. The experiment results show that hyperparameter fusion outperformed the average base learners in the area under the receiver operating characteristic (AUC) score. Moreover, removing the worst base learner can improve the AUC score for the ensemble. The discussion section predicts the worst base learner from correlations of normality rankings created by model outputs. The worst base learner has relatively small ranking correlations to the ensemble model compared to other base learners.

One-class classification (OCC) is a supervised classification problem where training data includes only one class. The goal is to classify data into one class or other classes. One of the solutions is a subtask-based approach; its idea is to use the error of an arbitrary subtask. This paper conceptualizes a novel subtask-based OCC approach using an OCC subtask (OCOCC), considered with a self-labeled dataset. The core hypothesis is that OCC subtask results for one class are better than those for other classes. The proposed framework allows a recursive process to provide endless OCC subtasks. Moreover, the OCOCC could create subtasks from all OCC algorithms, including state-of-the-art (SOTA). The OCOCC is experimented with image benchmark datasets, such as MNIST, Fashion MNIST, CIFAR10, and X-ray pneumonia. The OCC subtask is executed by applying existing OCC algorithms to a self-labeled dataset created by image rotations. The positive result is that OCOCC using OCAE subtask (OCOCAE) outperformed OCAE (OCC using autoencoder) in most datasets. Exploring the OCOCC framework is a promising direction for the future of OCC.

MNIST is a famous image dataset; several researchers evaluated their algorithms using MNIST and provided high accuracy. However, the accuracies were degraded on other datasets. Such an aspect raises the assumption that accuracy can be improved if all data were MNIST. Accordingly, this study proposes a preprocessing algorithm to transform all data into MNIST. In the proposal, an autoencoder (AE) is trained from MNIST, where the hypothesis lies that all decoder outputs are MNIST. Then, decoders are transferred to process feature vectors extracted from arbitrary input datasets. In the experiment, transformed data are compared with the original data in supervised classification. Although the accuracy is not improved, the proposed transformation method shows an advantage regarding privacy protection.

Autoencoder (AE) is a common technique for one-class classification (OCC). Reconstruction error (RE) is used to classify one seen class or other unseen classes. However, AE-based OCC (OCAE) does not provide a high AUC score. This study considers the hypothesis that RE is related to image entropy, and the OCAE is biased due to the image entropy differences. Based on such a hypothesis, this paper proposes image entropy equalization as the preprocessing technique. In which, image pixels are replaced by a defined set of pixels. Entropy equalized images are experimented with OCAE using MNIST, Fashion MNIST, and CIFAR10 datasets. Image Entropy Equalization improves AUC scores with several seen classes, where the improved classes have relatively high entropy on original images.

A Home Automation System (HAuSy) is a system of hardware and software which extends the traditional smart home architecture with a middle layer to increase its security and robustness. One of the added features is an automatic addressing when a new node is added to the system. The process of addressing wired nodes is straightforward. However, problems arise when large amounts of wireless Bluetooth Low Energy nodes must be dealt with. There are limitations that come from the hardware side, due to having three layers of architecture. It can be described as an evaluated bipartite graph. A proposal for how to deal with the address assignment problem is presented and the best fit is selected for future implementation. It is shown that complete task automation is not possible, and user interventions will be necessary. Finally, the technology requirements for the software plugin features so as to allow comfortably solving the presented issues are described.

Design of the Watch Dog health care application is presented in this paper. Solution is based on the client running on the smartphone and the server side running in the cloud. Sensors embedded in the smartphone are used for the measurement of the different information about the monitored person such as position, temperature, breath frequency etc. Basic algorithms evaluating current person's health status run on the smartphone. Measured data are sent to the second part of the application running in the cloud for deeper analysis. Suitability of the cloud solution for this application is discussed in this paper.