Ebook: Workshop Proceedings of the 19th International Conference on Intelligent Environments (IE2023)

Intelligent Environments (IE) refer to physical spaces that are enhanced by digital technologies and are designed to improve the quality of life of the people who live or work in them. These environments are equipped with technologies such as sensing systems and artificial intelligence that enable detection of changes in the environment, anticipate user requirements, and provide personalized services and experiences to users of such environments. The research community of Intelligent Environments explores core ideas of IEs in different contexts as well as critical issues needed to make them a reality, such as energy efficiency, computational constraints of edge devices and privacy issues. The 19th International Conference on Intelligent Environments (IE2023) was held in Mauritius between 27th June and 30th June 2023 and also in hybrid (online) mode for participants who could not come to the island. The following workshops are included this year’s conference:

- 12th International Workshop on the Reliability of Intelligent Environments (WoRIE’23)

- 3rd International Workshop on Artificial Intelligence and Machine Learning for Emerging Topics (ALLEGET’23)

- 3rd International Workshop on Self-Learning in Intelligent Environments (SeLIE’23)

- 2nd International Workshop on Edge AI for Smart Agriculture (EAISA’23)

- 12th International Workshop on Intelligent Environments Supporting Healthcare and Well-being (WISHWell’2023)

The dedication and hard work of the workshop chairs and committees have resulted in a high-quality collection of papers that represents the latest and most exciting developments in intelligent environments research. The papers contained within this volume represent a diverse array of cutting-edge research, and I am confident that they will inspire future breakthroughs in this exciting area of study. It is great to see the continued participation of the above workshops in the annual conference. With each passing year, some workshops, notably, WoRIE and WISHWell, have become an integral part of the conference program and it’s heartening to see the continued commitment of workshop organizers to share their expertise and contribute to the advancement of the respective fields. In addition, it is inspiring to see the continued success and growth of SeLIE, ALLEGET and EAISA to continue advancing research in their respective fields of IE and we are confident that these workshops will continue to have a lasting impact on the research community and beyond.

Finally, we would like to thank everyone who contributed in IE2023 for the commitment, for advancing research in the respective fields, and for sharing insights and expertise within the community. We also sincerely appreciate the support to our technical sponsors, the IEEE Systems Man & Cybernetics Society and IOS Press. We look forward to continuing working with you all as we push the boundaries of knowledge and explore new frontiers in Intelligent Environments.

Due to the rapid growth of smart grid applications all over the globe, it has become a more attractive target to malicious actors. Countries and stakeholders (e.g., governments) spend billions of dollars on ensuring the continuity and security of their smart grids for strategic and operational reasons. In fact, the risk associated with compromising a smart grid is considered among the highest in the cybersecurity world. This paper surveys a group of well-known smart grid intrusion detection datasets that are used in the development of machine learning-based intrusion detection systems. The study presents an analysis of these datasets and provides recommendations for researchers utilizing them.

The recent explosion of sensors enable our environment to act in an intelligent way. These Intelligent Environments rely on sense making of the sensors’ data streams. This process starts with reliable signal processing in real-time. This is challenging due to i) the low energy and computing resources of edge devices, ii) the signals’ non-stationary nature, and iii) the variety in software and hardware. To tackle this triplet of challenges, we present a WebAssembly-based hardware and software independent fast Continuous Wavelet Transform (fCWT), which excels in processing non-stationary signals at low costs. The application shows to be 2x-5.5x faster than competitors on speech, electrocardiogram (ECG), and vibration signals, enabling reliable real-time processing on edge devices. This yields new opportunities for the creation of safe and reliable Intelligent Environments.

Despite the wealthy of information biosignals cary, with Intelligent Environments (IE) they are often disregarded. We discuss issues we faced with integrating reliable biosignals into a real-world IE. These include the limited conductivity of dry sensors, movement artifacts, and placement issues. Subsequently, we introduce a real-time Signal Quality Indicator (SQI) for ElectroCardioGram (ECG), which consists of a Signal Loss Indicator (SLI) that detects signal capping, flatlining, high-frequency noise, and low-frequency noise. If the SLI detects a signal, the Signal Usability Indicator (SUI) subsequently processes the signal using the reference Pan-Tompkins algorithm and a dedicated filter to extract heart rate. The SQI marks what parts of the signal can and cannot be used for analysis. As such, it allows empirical calibration and, hence, the use of biosensors in real-world IE.

This paper describes a continuous air quality monitoring system using the LoRa standard. This network consists of two types of nodes: end nodes (sensor that measure humidity, temperature, carbon dioxide, particulate material) and gateway nodes (device that receives, processes, and sends the data to the cloud). The monitoring system was used in a real context in two points A and B, both located in the center of the city of Esmeraldas, Ecuador. Point A corresponds to a conventional house; while point B corresponds to a university. Two scenarios were experimented, that is, changing the physical conditions of the environment and changing the position of the nodes in terms of distance and altitude. Results allowed us to get a model for cities in developing countries to begin the process of transition to smart cities. We conclude that end nodes showed better communication if they were located at a high altitude above sea level. Therefore, the technologies used in the creation of the network fit the conditions of the environment and the urban space of the city of Esmeraldas —a city that is starting to implement technology to become a smart city in the future.

Sentiment analysis has become an indispensable tool in various domains, including e-business, e-commerce, e-tourism, e-mobility, e-governance, e-security, e-learning, and e-health. By analyzing public opinion and preferences, organizations and policymakers can make data-driven decisions to improve their products, services, and policies. The implementation of sentiment analysis across these domains not only enhances customer satisfaction and engagement but also contributes to the overall growth and development of these industries.

Twitter geolocation is useful for various purposes, including tracking COVID-19 perceptions, analyzing political trends, and managing natural disasters. However, accurately predicting geolocations based on tweet content remains a challenge. In the past, machine learning approaches have tried to solve this problem by training prediction models on previously seen data, but these models often struggle to generalize to unseen places. To overcome these limitations, in this work we present a framework based on Natural Language Processing (NLP), Knowledge Graphs (KG), and Semantic Web to find geographical entities on tweets’ content. KG facilitate the extraction of structured knowledge of texts in order to study their semantic analysis based on NLP techniques to search associated geographical coordinates to the entities of that KG; if there is explicit mention of places in the tweet, the Semantic Web is used to find geographical information associated with the entities present in the tweets’ content. To evaluate the precision of the prediction algorithm, we compare our predicted latitude and longitude coordinates with AlbertaT6 floods dataset. Our results show an F1 score up to 0.851 within a 10 kilometer radius.

The fast Continuous Wavelet Transform (fCWT) is used to improve Deep Convolutional Neural Networks (DCNN)’s Speech Emotion Recognition (SER). While being computationally efficient, the fCWT’s time-frequency analysis overcomes traditional methods’ resolution limitations (e.g., Short-Term Fourier Transform). fCWT-induced DCNNs are compared to state-of-the-art DCNN SER systems. Comparing different wavelet parameters, we also provide an empirical strategy for balancing temporal and spectral features in speech signals. We suggest that this strategy is of generic interest for non-stationary signal processing where large amounts of data are available. fCWT’s potential for improving SER accuracy in real-time applications is confirmed. In parallel, the variance in the cross-validation folds confirmed deep learning’s vulnerability on non-big data sets.

League of Legends (LoL) is a multiplayer online battle arena video game developed and published by Riot Games. It is a team-based game with over 140 characters to make epic plays with. The game blends the speed and intensity of an real-time strategy game (RTS) with role-playing game (RPG) elements. Two teams of powerful champions, each with unique designs and play styles, battle head-to-head across multiple maps and game modes. Exploratory data analysis (EDA) is a statistical technique that can be used to analyze this data to extract valuable information for both researchers and players. By using EDA techniques on LoL match data, players can identify patterns, trends, and relationships that can help optimize their gameplay strategy. EDA can also help players identify their strengths and weaknesses and important statistics for their gameplay. The paper provides an introduction to the treatment of LoL match data using EDA techniques. It presents the most common data analysis techniques and explores some examples of how to apply these techniques to LoL match data. Furthermore, the paper discusses some ways in which data analysis can help LoL players improve their game, such as identifying their strengths and weaknesses, patterns and trends, important statistics, and meta changes.

Transformer architectures and pre-training have facilitated building higher-capacity models and effectively utilizing this capacity for a wide variety of tasks. Various libraries consist of carefully engineered state-of-the-art Transformer architectures under a unified API. But still, there exists no such system which detects the chemicals present in the Beauty products reviews on online e-commerce websites. On the market, there are thousands of different cosmetic products, all with different combinations of ingredients. Methods such as token identification and entity recognition transformer model advancing in both model architecture and model pre-training are implemented to identify and categorize the named entities.

From the real-time forecasting of events to Visual analysis tasks, the state-of-the-art machine learning algorithms exhibit unmatched performance. Furthermore, with the ongoing traction of embedded computing, the deployment of machine learning algorithms on mobile devices is receiving increasing attention. There are numerous practical applications where hand-held devices having machine learning methods can be more useful due to their compact size and integrated resources. However, for the realization of ML methods on embedded devices, either the used algorithm should be less computationally complex or there should be some efficient way to implement a state-of-the-art algorithm on a less-powerful embedded device. In this paper, different approaches for reducing the computational complexity of a machine learning-based computer vision application are presented that can be helpful to make other such algorithms applicable on the embedded devices. Results show that the hardware architecture based exploitation can further improve the performance of an existing framework.

The intelligent transportation system (ITS) along with vehicular communications are making our daily life safer and easier e.g., saving time, traffic control, safe driving, etc. Many transmission mode selection and resource allocation schemes are mitigating to full the quality of service requirements with minimum latency and negligible interference. For message transmission to far away vehicles realistic cellular link and mode selection is a bottleneck, however, for nearby devices, the safety critical information needs V2V links. Reinforcement learning (RL) and Deep learning (DL) have reshaped vehicular communication in a new model where vehicles act like human beings and take their decisions autonomously without human intervention. In our work, we investigate Vehicle-to-vehicle (V2V) and Vehicle-to-infrastructure where each link takes a decision to find the optimal subband and power level. We investigated the case where each V2V link connected with Vehicle-to-Infrastructure (V2I) satisfying stringent latency constraints while minimizing interference. We exploit RL with DL to develop highly intelligent performance where agents effectively learn to select mode and share spectrum with V2V and V2I.

This paper introduces an Artificial Intelligence (AI)-enabled system to assist patients to follow a treatment plan at home. The deep learning model is a Convolutional Neural Network (CNN) classifier that is able to detect a drug even when shown in different orientations. The CNN model is trained for each patient based on his/her prescription medicine schedule. The advantage of the system is the dynamic functionality that makes it a good solution for personalized medication. The GUI demonstrates that the system can assist patients in taking the correct drug and prevent medication errors.

Coronavirus can lead to respiratory illnesses ranging from mild to severe, and even death, which makes early detection critical. However, current COVID-19 (Coronavirus Disease 2019) detection methods are not only expensive but also time-consuming. This poses a challenge, especially with an increasing number of patients and demand for testing kits. Waiting for test results for a few days is not ideal, as the outbreak can spread quickly in the meantime. To address this issue, we propose a COVID-19 prediction infrastructure using deep learning. This innovative android-based application uses a Convolutional Neural Network model, trained on a custom dataset with an accuracy of 97 percent, to predict whether COVID-19 is present or not. With this fast and low-cost approach, users can quickly detect COVID-19 and take appropriate actions to reduce the risk of transmission.