Sensors Data Processing Using Machine Learning.

Various sensors utilize computational models to estimate measured variables, and the generated data require processing [...].

A New Deep-Learning Method for Human Activity Recognition.

Currently, three-dimensional convolutional neural networks (3DCNNs) are a popular approach in the field of human activity recognition. However, due to the variety of methods used for human activity recognition, we propose a new deep-learning model in this paper. The main objective of our work is to optimize the traditional 3DCNN and propose a new model that combines 3DCNN with Convolutional Long Short-Term Memory (ConvLSTM) layers. Our experimental results, which were obtained using the LoDVP Abnormal Activities dataset, UCF50 dataset, and MOD20 dataset, demonstrate the superiority of the 3DCNN + ConvLSTM combination for recognizing human activities. Furthermore, our proposed model is well-suited for real-time human activity recognition applications and can be further enhanced by incorporating additional sensor data. To provide a comprehensive comparison of our proposed 3DCNN + ConvLSTM architecture, we compared our experimental results on these datasets. We achieved a precision of 89.12% when using the LoDVP Abnormal Activities dataset. Meanwhile, the precision we obtained using the modified UCF50 dataset (UCF50mini) and MOD20 dataset was 83.89% and 87.76%, respectively. Overall, our work demonstrates that the combination of 3DCNN and ConvLSTM layers can improve the accuracy of human activity recognition tasks, and our proposed model shows promise for real-time applications.

Physical and Digital Infrastructure Readiness Index for Connected and Automated Vehicles.

In this paper, we present an assessment framework that can be used to score segments of physical and digital infrastructure based on their features and readiness to expedite the deployment of Connected and Automated Vehicles (CAVs). We discuss the equipment and methodology applied for the collection and analysis of required data to score the infrastructure segments in an automated way. Moreover, we demonstrate how the proposed framework can be applied using data collected on a public transport route in the city of Zilina, Slovakia. We use two types of data to demonstrate the methodology of the assessment-connectivity and positioning data to assess the connectivity and localization performance provided by the infrastructure and image data for road signage detection using a Convolutional Neural Network (CNN). The core of the research is a dataset that can be used for further research work. We collected and analyzed data in two settings-an urban and suburban area. Despite the fact that the connectivity and positioning data were collected in different days and times, we found highly underserved areas along the investigated route. The main problem from the point of view of communication in the investigated area is the latency, which is an issue associated with infrastructure segments mainly located at intersections with heavy traffic or near various points of interest. The low accuracy of localization has been observed mainly in dense areas with large buildings and trees, which decrease the number of visible localization satellites. To address the problem of automated assessment of the traffic sign recognition precision, we proposed a CNN that achieved 99.7% precision.

An Improved IoT-Based System for Detecting the Number of People and Their Distribution in a Classroom.

This paper presents an improved IoT-based system designed to help teachers handle lessons in the classroom in line with COVID-19 restrictions. The system counts the number of people in the classroom as well as their distribution within the classroom. The proposed IoT system consists of three parts: a Gate node, IoT nodes, and server. The Gate node, installed at the door, can provide information about the number of persons entering or leaving the room using door crossing detection. The Arduino-based module NodeMCU was used as an IoT node and sets of ultrasonic distance sensors were used to obtain information about seat occupancy. The system server runs locally on a Raspberry Pi and the teacher can connect to it using a web application from the computer in the classroom or a smartphone. The teacher is able to set up and change the settings of the system through its GUI. A simple algorithm was designed to check the distance between occupied seats and evaluate the accordance with imposed restrictions. This system can provide high privacy, unlike camera-based systems.

A New Approach for Abnormal Human Activities Recognition Based on ConvLSTM Architecture.

Recognizing various abnormal human activities from video is very challenging. This problem is also greatly influenced by the lack of datasets containing various abnormal human activities. The available datasets contain various human activities, but only a few of them contain non-standard human behavior such as theft, harassment, etc. There are datasets such as KTH that focus on abnormal activities such as sudden behavioral changes, as well as on various changes in interpersonal interactions. The UCF-crime dataset contains categories such as fighting, abuse, explosions, robberies, etc. However, this dataset is very time consuming. The events in the videos occur in a few seconds. This may affect the overall results of the neural networks that are used to detect the incident. In this article, we create a dataset that deals with abnormal activities, containing categories such as Begging, Drunkenness, Fight, Harassment, Hijack, Knife Hazard, Normal Videos, Pollution, Property Damage, Robbery, and Terrorism. We use the created dataset for the training and testing of the ConvLSTM (convolutional long short-term memory) neural network, which we designed. However, we also test the created dataset using other architectures. We use ConvLSTM architectures and 3D Resnet50, 3D Resnet101, and 3D Resnet152. With the created dataset and the architecture we designed, we obtained an accuracy of classification of 96.19% and a precision of 96.50%.

Vehicle Classification Based on FBG Sensor Arrays Using Neural Networks.

This article is focused on the automatic classification of passing vehicles through an experimental platform using optical sensor arrays. The amount of data generated from various sensor systems is growing proportionally every year. Therefore, it is necessary to look for more progressive solutions to these problems. Methods of implementing artificial intelligence are becoming a new trend in this area. At first, an experimental platform with two separate groups of fiber Bragg grating sensor arrays (horizontally and vertically oriented) installed into the top pavement layers was created. Interrogators were connected to sensor arrays to measure pavement deformation caused by vehicles passing over the pavement. Next, neural networks for visual classification with a closed-circuit television camera to separate vehicles into different classes were used. This classification was used for the verification of measured and analyzed data from sensor arrays. The newly proposed neural network for vehicle classification from the sensor array dataset was created. From the obtained experimental results, it is evident that our proposed neural network was capable of separating trucks from other vehicles, with an accuracy of 94.9%, and classifying vehicles into three different classes, with an accuracy of 70.8%. Based on the experimental results, extending sensor arrays as described in the last part of the paper is recommended.

A Smart IoT System for Detecting the Position of a Lying Person Using a Novel Textile Pressure Sensor.

Bedsores are one of the severe problems which could affect a long-term lying subject in the hospitals or the hospice. To prevent lying bedsores, we present a smart Internet of Things (IoT) system for detecting the position of a lying person using novel textile pressure sensors. To build such a system, it is necessary to use different technologies and techniques. We used sixty-four of our novel textile pressure sensors based on electrically conductive yarn and the Velostat to collect the information about the pressure distribution of the lying person. Using Message Queuing Telemetry Transport (MQTT) protocol and Arduino-based hardware, we send measured data to the server. On the server side, there is a Node-RED application responsible for data collection, evaluation, and provisioning. We are using a neural network to classify the subject lying posture on the separate device because of the computation complexity. We created the challenging dataset from the observation of twenty-one people in four lying positions. We achieved a best classification precision of 92% for fourth class (right side posture type). On the other hand, the best recall (91%) for first class (supine posture type) was obtained. The best F1 score (84%) was achieved for first class (supine posture type). After the classification, we send the information to the staff desktop application. The application reminds employees when it is necessary to change the lying position of individual subjects and thus prevent bedsores.