Bachelor and Master projects under my supervision are both research and application-oriented, yet primarily related to the topics we are currently investigating in our group. You can get a sense by having a look at some of the completed projects (see below). Do not hesitate to contact me if you are interested in a project or in being involved in the activities of our group. Research internships are also a possibility, for example through the Erasmus+ program, so please send me an email if you are interested.
I am also looking for passionate students interested in pursuing a PhD in Data Science, Ubiquitous Computing, and Digital Health in our group. If you are interested please contact me by email or drop by my office in order to discuss possibilities for funding your studies.
I am happy to discuss postdoc opportunities in our group, also potentially through fellowships such as those offered by the EU Marie Skłodowska-Curie Actions or the ERCIM Program.
* Thesis awarded with Cum Laude honors.
Despite the popularity of human behavior recognition systems, most available solutions are developed to operate on predefined settings and fixed sensor setups. Real-world behavior recognition applications and users demand more flexible sensor configurations, which may deal with potential adverse situations such as defective or missing sensors. This thesis aims at exploring new ways for supporting seamless and reliable behavior recognition. This work revolves around the hypothesis that ontologies are the perfect means to comprehensively describe the available resources that could be utilized for human behavior recognition in the wild. Not only do ontologies provide interoperability, but this semantic representation of knowledge also enables, through ontological reasoning, the smart selection of the best resources that could provide behavior recognition capabilities in real world scenarios.
Lower limb injuries are the most common injuries among sport practitioners, specially those related to the knee and ankle joints. Consequently, it has appeared a growing interest in the identification of the risk factors for those injuries, leading to the common objective of predicting the injuries and avoid their appearance. This work presents a new marker tracking algorithm based on computer vision techniques for measuring knee injuries. The system has been tested on a professional football team and the data acquired have been analyzed using data mining techniques, in order to establish a first approach to a classification of the players in three levels of injury risk. It is also determined the most informative injury risk factors for its use in future analyses.
Breast cancer is the most common tumor in western women and statistically 1 out of 8 women will develop breast cancer over their lifetime. The rehabilitation the patient should follow after surgery is critical to recover from the suffered illness. In this work, a system composed of 3 applications, one for smartwatches, one for smartphones and a web application, is presented. Applications for handheld devices are directed to the patient who is undergoing rehabilitation and allow to monitor parameters of interest that will indicate whether the rehabilitation process being followed is improving the health of the patient. The web application is directed to a medical expert with the objective of tracking rehabilitation conducted by the patients.
* Project awarded with the Talentum Startup Grant (Telefonica, Spain) and the second prize in the VI UGR Entrepreneurial Contest (University of Granada, Spain).
This work extends a prior system for the automatic estimation of the trunk endurance using mobile and wearable sensors. Particularly, a new cloud-based storage and processing is proposed for the persistence and analysis of the data registered during the test sessions. Moreover, data mining techniques are implemented to identify correlations between the user activity, self-assessment and the endurance results. In addition, to verify the usefulness of the system, the endurance tests have been tested by the new developed system for a professional football team. This study especially aims to be the start point for the development of a new classification system of muscle activity for the lower back, based on the use of clustering and data mining in the new data generated by the application.
Tons of datasets are increasingly available in many repositories, e.g., DATA.GOV, UCI or AWS. However, there is a clear lack of webtools that help automate the process of analyzing those datasets. This thesis contributes with a new user-friendly webtool especially oriented to support researchers in the analysis of existing and newly collected datasets. The website allows users to upload their datasets and share them with the community. The uploaded data can be analyzed through various methods and functions. Both raw and processed data can be represented through advanced visualization techniques. The tool implements several algorithms to create machine learning models with the uploaded data. Based on these generated models new data can be classified or predicted.
* Project awarded with honors and with the ETSIIT second prize to the best project in Telecommunications Engineering (University of Granada, Spain).
Low back pain is a primary cause of disability worldwide. Trunk endurance tests are normally considered to assess the muscle status; however, traditional procedures suffer from practical limitations leading to inaccurateness. This work developed mDurance, an innovative mHealth system to support specialists in the trunk endurance assessment. A wearable MIMU sensor is used to dynamically track the patient trunk posture to estimate the test duration, while portable EMG sensors are employed to seamlessly measure the muscle stress. The information is managed by an app facilitating the expert routine and minimizing human errors. The mDurance potential is shown through a case study. In order to show the potential of the mDurance system, a case study has been conducted. The results of this study prove the reliability of mDurance and further demonstrate that practitioners are certainly interested in the regular use of a system of this nature.
Activity recognition models are normally based on predefined on-body sensor positioning. However, the sensor deployment may vary due to several conditions. User self-attachment, firmness of the attachment (loose of fitting) or displacements due to the use of the sensors may introduce variations with respect to the original setup. Furthermore, the accuracy of the recognition system may strongly depend on the particular body position considered for the mounting of the sensor. This project explores the effects that different sensor configurations/positions may have on the system recognition capabilities. Furthermore, this thesis presents new techniques that attempt to autonomously identify where a sensor is on-body located. This information is particularly interesting to adapt the recognition methods to the best sensor configuration at each time.
* Project awarded with the Emilio Herrera Linares award to the best Spanish graduate research project (University of Granada, Spain).
Mobile health is an emerging field which is attracting much attention. Nevertheless, tools for the development of mobile health applications are lacking. This work presents mHealthDroid, an open source Android implementation of a mHealth Framework designed to facilitate the rapid and easy development of biomedical apps. The framework is devised to leverage the potential of mobile devices like smartphones or tablets, wearable sensors and portable biomedical devices. The framework provides functionalities for resource and communication abstraction, biomedical data acquisition, health knowledge extraction, persistent data storage, adaptive visualization, system management and value-added services such as intelligent alerts, recommendations and guidelines.
* Project awarded with the Emilio Herrera Linares award to the best Spanish graduate research project (University of Granada, Spain).
The delivery of healthcare services has experienced tremendous changes during the last years. Mobile health or mHealth is a key engine of advance in the forefront of this revolution. Although there exists a growing development of mobile health applications, there is a lack of tools specifically devised for their implementation. This work presents mHealthDroid, an open source Android implementation of a mHealth Framework designed to facilitate the rapid and easy development of mHealth and biomedical apps. The framework is particularly planned to leverage the potential of mobile devices such as smartphones or tablets, wearable sensors and portable biomedical systems. These devices are increasingly used for the monitoring and delivery of personal health care and wellbeing. An exemplary application is also presented along this work to demonstrate the potential of mHealthDroid. This app is used to investigate on the analysis of human behavior, which is considered to be one of the most prominent areas in mHealth. An accurate activity recognition model is developed and successfully validated in both offline and online conditions.
Signal segmentation is a crucial stage in the activity recognition process; however, this has been rarely and vaguely characterized so far. Windowing approaches are normally used for segmentation, but there exists no clear consensus on which window size should be preferably employed. This work investigates the effects of the windowing procedure on the activity recognition process. To that end, diverse recognition systems are tested for several window sizes also including the values used in previous works. The study proves that reduced window sizes lead to a better recognition of the activities, which goes against the generalized idea of using long data windows. Moreover, this work explores the combination of multiple recognizers operating on different window sizes to optimize the recognition performance.
The ever progressing technological advances on the development of mobile phones, medical sensors and wireless communication systems are supporting a new generation of unobtrusive, portable and ubiquitous health monitoring systems for a continuous patient assessment and a more personalized health care. In this paper we present PhysioDroid, an Android based application operated together with a wearable monitor capable of measuring vital information such as electrocardiogram (ECG), heart rate (HR), respiration rate (BR), skin temperature and motion directly from the human body. The application provides gathering, storage and processing features for the body sensor data. Likewise, the system works as a gateway enabling data transfer to a remote server which may be used for further processing and analysis. PhysioDroid also implements visualization of physiological information mainly intended to trigger alerts and emergency calls when abnormalities or risk situations are detected.
There exist a growing number of mobile apps in the health domain; however, little contribution has been specifically provided, so far, to operate this kind of apps with wearable physiological sensors. This work contributes to the development of a new medical application that provides a personalized means to remotely monitor and evaluate users’ conditions. The system provides ubiquitous and continuous vital signs analysis, such as electrocardiogram, heart rate, respiration rate, skin temperature, and body motion, intended to help empower patients and improve clinical understanding. Based on the analysis of the user's vital sign and behavioral information the application can trigger personal alerts when abnormal situations are detected.
Although much effort has been put in the development of valid activity recognition models, most related works do not regularly consider the preprocessing of the motion data. This work evaluates the effect of processing human inertial-sensing signals for diverse daily living activities. The findings of this study demonstrate that depending on the target activity set different preprocessing techniques should be considered.
The use of smartphones for the recognition of human activities has attracted much interest during the recent years. Smartphones are at the reach of most consumers, and as such, these devices are the perfect means for the development of realistic activity recognition applications. This work explores the use of the inertial sensors provided by standard smartphone devices to identify some of the most commonplace user activities.
Physical activity recognition is an area of growing interest given its multiple application domains. Although there exist several studies analyzing the potential capabilities of diverse recognition systems, there are few contributions exploring the use of smartphones for determining the user activity. A new application is developed to help recognizing the daily living physical activity of the user by using the available sensors in the smartphone. The application can collect lots of data from users all over the world. These data can be uploaded to a server to be analyzed in the future. With the help of this monitoring system we can find out more about the physical activity of the users and increase the accuracy of the application.
With the rapidly ageing population, mental disorders associated to ageing are becoming more prevalent. Cognitive impairments, such as dementia, can cause severe problems in daily life. Currently diagnosis of cognitive impairment is performed through medical assessments after potential symptoms have been detected. Digital approaches can provide more immediate and continuous assessment which could allow for much earlier diagnosis of mental disorders. This work develops and explores the use of smartphone-based cognitive experience sampling techniques based on the digitalisation of standardised clinical cognitive assessment tests.
Autistic people express their emotions in a special way, which makes it hard for others to interpret them correctly. This is especially troublesome for autistic children, whose parents have difficulty reading their emotions. The different expression of emotions by autistic people influences affective signals, which makes technology based on facial recognition, body language or voice intonation unreliable. Sensors for these affective signals can be placed in a variety of products, but these products should adhere to specific design guidelines based on user requirements of the target group, which is sensitive to stimuli and has difficulty adapting. This work elaborates on this user requirements by creating a set of guidelines for designing sensors to measure affective signals from autistic children.
Twenty one million bicycles, including electric bicycles, are sold every year in Europe. A bicycle is a useful vehicle that can be used by all age groups. However, it appears that elderly cyclists have a higher risk of getting injured in traffic. Statistics show that elderly people have a very large risk of getting involved in a cycling accident as compared to other age groups. In this research the technical possibilities of smartphones and mobile sensors are explored with the aim of helping elderly cyclists in traffic.
Staff working in healthcare and retirement homes frequently face high workloads and accordingly high stress levels are reported by healthcare personal. In hospitals and nursing homes valuable time is lost while searching for medical devices and supporting aids like lifts or beds. Another disadvantage of lost equipment is that the required maintenance is not conducted at the time of need. A system that easily tracks and finds such devices would save valuable time and accordingly would reduce the work stress of healthcare staff and improve the quality of work. This thesis describes the feasibility study of a simple and low cost tracking system for medical devices based on existing IT infrastructure available in every hospital and retirement homes nowadays: A Wi-Fi network. An indoor Wi-Fi fingerprint system was developed, tested and evaluated. This study has shown that a tracking system based on Wi-Fi position is feasible.
* Project awarded with honors (University of Granada, Spain).
Knee alignment measurements are one of the most extended indicators of knee-complex injuries such as anterior cruciate ligament injury and patellofemoral pain syndrome. The Frontal Plane Projection Angle (FPPA) is widely used as a 2-D estimation of knee alignment. However, traditional procedures to measure this angle suffer from practical limitations, which lead to huge time investments when evaluating multiple subjects. This thesis presents a novel video analysis system aimed at supporting experts in the dynamic measurement of the FPPA in an cost-effective and easy way. The system employs the Kinect V2 depth sensor to track reflective markers attached to the patient leg joints to provide an automatic estimation of the angle formed by the hip, knee and ankle joints. Information registered by the sensor is processed and managed by a computer application that simplifies the expert’s work and expedites the analysis of the test results.
The goal of this project was to enable a larger patient group to exploit the gait rehabilitation pressure sensitive LED floor “Black Spinel”. This project involves the integration of a crutch into the Black Spinel. A prototype including a force sensitive resistor and Bluetooth module has been developed. The prototype was tested on the Black Spinel. A clear distinction was seen when steps were made on the floor as well as on the crutch. Future research might further explore the addition of a second crutch, improve calibration and filter out phantom measurements.
* Project awarded with a Graduate Research Assistantship (University of Granada, Spain).
The evaluation of cancer patients’ recovery is still subject to great subjectivity from a clinical perspective. Many different systems have been successfully implemented for physical activity evaluation, nonetheless there is still a big leap into performance status evaluation with ECOG and Karnofsky’s, which are possibly the most widely-used performance status scores. This thesis presents a novel system for automatically recording the patient's activity and derive these scores by using a smartphone and smartwatch. A gamification approach has also been designed for increasing patients’ motivation in their recovery. Furthermore, unprecedented algorithms for performance status and physical activity assessment have been developed to help oncologists in their diagnoses.