A5036773239- Context-Aware Activity Recognition Systems
- Energy Efficient Wireless Sensor Networks
- Human Mobility and Location-Based Analysis
- Gait Recognition and Analysis
- COVID-19 epidemiological studies
- COVID-19 Pandemic Impacts
- Human Pose and Action Recognition
- Anomaly Detection Techniques and Applications
- Physical Activity and Health
- Mobile Crowdsensing and Crowdsourcing
- Sports Analytics and Performance
- Artificial Intelligence in Games
- Green IT and Sustainability
- Innovative Human-Technology Interaction
- Obesity, Physical Activity, Diet
- Educational Games and Gamification
- SARS-CoV-2 and COVID-19 Research
- Mobile Health and mHealth Applications
- Artificial Intelligence in Healthcare
- Smart Grid Energy Management
- Water Quality Monitoring Technologies
- IoT-based Smart Home Systems
- User Authentication and Security Systems
- COVID-19 diagnosis using AI
- Non-Invasive Vital Sign Monitoring
Jožef Stefan Institute
2014-2023
Jožef Stefan International Postgraduate School
2017-2020
The Sussex-Huawei Locomotion-Transportation recognition challenge presents a unique opportunity to the activity-recognition community - providing large, real-life dataset with activities different from those typically being recognized. This paper describes our submission (team JSI Classic) competition that was organized by authors. We used carefully executed machine learning approach, achieving 90% accuracy classifying eight (Still, Walk, Run, Bike, Car, Bus, Train, Subway). first step data...
In recent years, activity recognition (AR) has become prominent in ubiquitous systems. Following this trend, the Sussex-Huawei Locomotion-Transportation (SHL) challenge provides a unique opportunity for researchers to test their AR methods against common, real-life and large-scale benchmark. The goal of is recognize eight everyday activities including transit. Our team, JSI-Deep, utilized an approach based on combining multiple machine-learning following principle knowledge. We first created...
One key task in the early fight against COVID-19 pandemic was to plan non-pharmaceutical interventions reduce spread of infection while limiting burden on society and economy. With more data being generated, it became possible model both trends intervention costs, transforming creation an into a computational optimization problem. This paper proposes framework developed help policy-makers best combination change them over time. We hybrid machine-learning epidemiological forecast trends,...
The Sussex-Huawei Locomotion Challenge 2019 was an open competition in activity recognition where the participants were tasked with recognizing eight different modes of locomotion and transportation. main difficulty challenge is that training data recorded a smartphone placed body location than test data. Only small validation set all locations provided to enable transfer learning. This paper describes our (team JSI First) approach, which we first derived additional sensor streams from...
The SHL recognition challenge 2020 was an open competition in activity where the participants were tasked with recognizing eight different modes of locomotion and transportation smartphone sensors. main challenges that training data recorded by a person than validation test data, location unknown to participants. We, team "Third time's charm", tackled first attempting identify persons clustering, then performed cluster/person-specific feature selection build separate classifier for each...
We present the e-Gibalec system, designed to encourage schoolchildren towards a more active lifestyle. The system consists of mobile application that, through sensors built into smartphone, detects children's physical activity and rewards them in game-like manner. It also web that allows parents education teachers look at history, so they can further motivate if needed. discuss motivational mechanisms employed provide an evaluation accuracy activity-recognition component, pilot study...
This paper presents a novel method for activity recognition and estimation of human energy expenditure with smartphone an optional heart-rate monitor. The detects the presence devices, normalizes orientation phone, its location on body, uses location-specific models to recognize estimate expenditure. normalization detection significantly improve accuracy; estimated is more accurate than that provided by state-of-the-art dedicated consumer device.
The recognition of high-level activities (such as work, transport and exercise) with a smartphone is poorly explored topic. This paper presents an approach to such activity that relies on the user's location, physical activity, ambient sound other features extracted from sensors. It works in user-independent fashion, but can also take advantage labeled by user. was evaluated real-life dataset consisting ten weeks recordings. While most were recognized quite accurately, some revealed two...
The COVID-19 pandemic affected the whole world, but not all countries were impacted equally. This opens question of what factors can explain initial faster spread in some compared to others. Many such are overshadowed by effect countermeasures, so we studied early phases infection when countermeasures had yet taken place. We collected most diverse dataset potentially relevant and metrics date for this task. Using it, show importance different factor categories as determined both statistical...
We present the "e-Gibalec" system, which was designed to stimulate and motivate schoolchildren be physically active assist physical education (PE) teachers parents track their progress. The system consists of a smartphone application uses built-in sensors monitor activities, web for PE parents. an overview its functionality. also provide initial evaluation accuracy monitoring obtained on 10 schoolchildren, subjective mobile application, is quite positive.
From 2018 to 2021, the Sussex-Huawei Locomotion-Transportation Recognition Challenge presented different scenarios in which participants were tasked with recognizing eight modes of locomotion and transportation using sensor data from smartphones. In 2019, main challenge was one location recognize activities sensors another location, while following year, person other persons. We use these two as a framework analyze effectiveness components machine-learning pipeline for activity recognition....
The recognition of the user's context with wearable sensing systems is a common problem in ubiquitous computing. However, typically small battery such often makes continuous impractical. strain on can be reduced if sensor setting adapted to each context. We propose method that efficiently finds near-optimal settings for It uses Markov chains simulate behavior system different configurations and multi-objective genetic algorithm find set good non-dominated configurations. was evaluated three...
The recognition of the user's context with wearable sensing systems is a common problem in ubiquitous computing. However, typically small battery such often makes continuous impractical. An efficient method to reduce strain on employ duty-cycling – periodically turning sensors and off. Its benefits increase if duration those periods tailored each context. In this work we present general mathematical model predict effect different duty-cycle parameters system discuss ways selecting suitable...
Detection of the user's context with mobile sensing systems is a common problem in ubiquitous computing. However, typically small battery such often making continuous detection impractical. The strain on can be reduced if sensor setting adapted to each context. We propose method that efficiently finds near-optimal settings. It uses Markov chains simulate behaviour system different configurations, and multi-objective genetic algorithm find set good non-dominated configurations.
Context recognition using wearable devices is a mature research area, but one of the biggest issues it faces high energy consumption device that sensing and processing data. In this work we propose three different methods for optimizing its use. We also show how to combine all further increase savings. The by adapting system settings (sensors used, sampling frequency, duty cycling, etc.) both detected context directly sensor This done mathematically modeling influence multiobjective...
An important problem in the field of context recognition is preserving battery life sensing device. In this work we introduce cost-sensitive learning that takes into account cost having sensors active and tries to balance it against classification error. The benefits can be amplified by adapting classifier each context. This approach was tested on two real-life datasets where achieved 78% 80% reduction energy consumption exchange for almost no accuracy loss.