Master and Bachelor semester projects

Fall 2021


Orthopedics and sport

Contact person: Mahdi Hamidi Rad

Swimming analysis with IMU sensors has got more and more interesting for researchers recently because IMUs are more functional than video-based systems under water. Detection of the stroke phases and finding their duration is one of the analyses done for evaluation of swimmers strokes. Many studies focused on using accelerometer and gyroscope data for phase detection during strokes, the shortcoming of which is that the data is not sensitive to change of medium (hand entering or exiting the water). The idea of this project is to find these two events in different stroke styles using pressure data, obtained from an encapsulated barometer attached to swimmer’s wrist. The data for 17 swimmers is available in different stroke styles and the student should develop algorithms to detect hand entering and exiting water in different stroke styles. Swimmers’ videos are available for validation.

Skill required: Matlab programming, knowing the basics of signal processing.

Contact person: Mahdi Hamidi Rad

Velocity is one of the most important metrics for swimmer’s performance evaluation. Swimming coaches are interested in velocity in different swimming phases from wall to wall, in order to assess each phase separately. The focus of this project is on the start phase of swimming which includes wall push-off and glide. The maximum velocity during wall push-off (representing push-off strength) and the velocity at the end of glide (showing how much the swimmer’s body stayed streamlined during the glide) are to be estimated using sacrum IMU data. The data (acceleration and angular velocity) for 19 swimmers is available in four swimming techniques synchronized with validated velocity data (obtained from a speedometer) during start. The student should adopt machine learning approaches to estimate these two velocity values using sacrum data.

Skill required: Fluency in English, Matlab programming, familiar with machine learning.

Contact person: Salil Apte

The overall goal of the project is to investigate the evolution of stride-to-stride variability during a long distance run and explore its value in improving the understanding of the changes that are entailed by fatigue. Biomechanical parameters such as stride time, foot strike, running speed, contact time, etc. will be provided based on a dataset of 15 half-marathon participants. First step will be to understand the behaviour of these parameters using the standard deviation, interquartile range, detrended fluctuation analysis, multiscale entropy, and power spectral density. Second step will be to investigate the influence of fatigue on the evolution of these metrics. A better knowledge of the individual evolution of these parameters can supplement the knowledge produced by functional tests and improve training to optimize capacities and prevent injuries.

Skill required: Basic statistical analysis, basics of Signal processing using Matlab, good grasp of English, and an interest in sports (especially running).

 

Contact person: Salil Apte

The overall goal of the project is to investigate the evolution of the trunk movement during a long distance run and explore its value in improving the understanding of the influence of acute fatigue. However, exploration is needed to understand which metrics are useful to investigate the trunk movement and detect movement compensations while running. Examples of such metrics include entropy and dynamic stability measures, etc. First step will be to estimate these parameters from the provided trunk inertial sensor data and the second step will be to investigate the influence of fatigue on the evolution of these metrics. Bonus step would be the estimation of trunk lean angle during the run. A better knowledge of the individual evolution of these parameters can supplement the knowledge obtained from the foot inertial sensors and improve training to optimize capacities and prevent injuries.

Skill required: Basic statistical analysis, basics of Signal processing using Matlab, good grasp of English, and an interest in sports (especially running).


Instrumentation

No projects at the present time


Gait Analysis

Contact person: Mina Baniasad

Project in collaboration with Sharif University of Technology (Prof. Farzam Farahmand)

Nowadays, planning of the surgeries for treatment of children suffering from cerebral palsy is mainly based on the gait analysis whether observational or instrumented. A gait cycle consists of a series of events that sequentially occur including foot contact and foot off. Based on these events in both legs, many tempo-spatial parameters can be extracted which are clinically important. Furthermore, observational gait analysis methods consider these moments for evaluation of the patients and by automatically identifying these moments, we would be able to provide the required information for the clinicians. Several algorithms have already been developed for normal gait. However, since the pattern of walking alteres in patients with neuromuscular diseases, these algorithms are not able to detect the events in patients. Currently, an expert defines the events based on the marker position and force plate data which is time consuming. The main of this project is to automatically identify the events in patients with crouch gait using only the 3D position of foot markers. Data required for this project has already been collected using vicon camera and two force plates.

Skill required: Matlab programming

Contact person: Mina Baniasad

Medial knee osteoarthritis is one of the primary causes of disability and pain in the elderly. The moment we applied to our knees during walking in the frontal plane has been positively associated with the presence, severity, progression, and pain of osteoarthritis. Several methodologies existed in the literature, and the main aim of this project is to compare the characteristics of knee adduction moment using different methods in inverse dynamics.

A comprehensive data, including motion capture (Vicon), force plate, and EMG has already been collected from 26 healthy subjects. Analyses will include investigating four methodologies for inverse dynamics and finally compare the characteristics of the knee adduction moment.

Skill required: Matlab programming, Dynamics

Contact person: Mina Baniasad

Gait analysis is widely used to provide practical hints for clinicians to treat the patients better. One of the limitations regarding the gait analysis is that each lab requires its own normative data, which would be very time consuming and expensive. Applying different methodologies to extract the clinical parameters is one of the sources of the differences between labs. So, the scheme of this project is to investigate the effect of one source of variations.

It has been shown that the moment applied to the knee in the frontal plane is strongly correlated with the level of knee pain. To estimate the knee adduction moment, first, we need to approximate the hip and knee joint center. In this regard, several methods have been proposed, including functional calibration and anthropometric data. The main aim of this project is to compare the results of these two methods in different anatomical planes and different phases of the gait cycle and finally to investigate its effect on knee adduction moment. The required data has already been collected from 26 healthy subjects.

Skill required: Matlab programming

Contact person: Anisoara Ionescu

Freezing  of  gait  (FOG),  is  defined  as  a  “brief, episodic absence or marked reduction of forward progression of  the  feet  despite  the  intention  to  walk” . FOG events typically manifest as a sudden and transient inability to move, typically occurring on initiating gait, on turning while walking, or when experiencing stressful situations. FOG can lead to fall, reduced mobility and decreased quality of life. Automatic detection and characterization of FOG in real-life situations using wearable inertial sensors could provide clinicians with valuable information that can be used to improve the treatment.

The objective of this project is to use existing database(s) recorded with multiple body-worn inertial sensors and to implement and evaluate the most promising state of the art methods for FOG detection and characterization.

Skill required: Matlab programming, machine learning, signal processing


Rehabilitation

No projects at the present time


Activity Monitoring

Contact person: Anisoara Ionescu

These days, when you look at the market of electronic gadgets, you can easily find a large number of different wrist-based activity trackers, which all are trying to automatically monitor and assess human physical activities. However, most of them employ GPS sensors, especially for challenging activities like cycling. The main problems of GPS is its high power consumption as well as the essential need to communicate with satellites, which can be easily discarded in urban environments. Consequently, in this project, we will pursue a novel idea to devise an algorithm only based on low-power and independent inertial (Accelerometer and gyroscope) and barometric pressure sensors, attached to the wrists, to automatically detect cycling periods during daily life. The project will also contain a few simple measurements for algorithm development.

Skill required: Fluency in English, MATLAB programming, some experiences in signal processing and machine learning.

Contact person: Anisoara Ionescu

How to characterize and quantify the response of physiological systems to physical, psychological and disease-related stressors is an important question in clinical research. Wearable sensor technologies allow recording of continuous multimodal data (e.g., ECG, respiration, blood pressure, body movement, etc) in real-life conditions providing opportunity to characterize and quantify the temporal/dynamic relationship between various physiological signals.  The objective of this study is to investigate analytical methods for temporal causality analysis of time-series describing fluctuations of physiological signals. The analyses will be conducted on existing databases including multimodal sensor recordings.

Skill required: Matlab programming, signal processing

Contact person: Salil Apte

The main purpose of this project is to help people in developing and retaining hand hygiene habits, especially in the context of the current Covid-19 situation. The platform will be composed of a wrist-worn accelerometer sensor, communicating with a smartphone via bluetooth. Machine learning (ML)  algorithms to detect the two activities, hand-wash and face-touch, have already been developed and validated. Within this project, the student will implement the said algorithms on the smartwatch accelerometer signals and develop a smartphone application to communicate with the smartwatch. Secondary work can include the exploration of the received signal strength indicator (RSSI) for the bluetooth communication and its potential to improve the ML algorithm. The overall goal is to provide feedback to users in the form of reminders for washing hands and vibration alerts for face touching, in order to improve their hygiene habits. The project will also involve some simple data collection protocols for refining and testing the application.

Skill required: Essentials of application development for Android/iOS, basic idea about bluetooth communication, and some knowledge about machine learning

Contact person: Gaëlle Prigent

Subjective Vitality is one component under the umbrella of well-being and describes the conscious experience of possessing energy and aliveness. It triggers the activation process to initiate actions and can be a very important factor for exercise maintenance. In this research, the focus is on the vitality-exercise relationship (specifically running). To verify the theory of running intensity and vitality – acute and long-term effects, we are conducting an 8 weeks running intervention study with follow-up measurements. Participants are equipped with electrocardiogram and accelerometer sensors during one week before and after the intervention to measure their physical activity (PA) behaviour, sleep quality and heart rate variability throughout daily activities. 

With the advancement of wearable sensing technology, continuous monitoring of daily physical activity (PA) has brought in new insights for human behavior analysis. Variance in day-to-day activities and their pattern of changes over time are of interest to characterize individuals’ specific behavior and predict the trend of functional status, especially in ageing population. State-of-the-art clinical-oriented analyses identified key components of PA assessment as the type, duration and intensity of the activities. The originality of this work is to enrich the above-mentioned PA assessment by adding physiological information. Combining PA and heart rate variability (HRV) throughout daily activities is a promising approach to assess vital signs and help understanding the effects of practicing intense exercise on well-being.  

The objectives of the project are to: (1) analyze the PA patterns in data collected PRE/POST intervention using methodology developed in LMAM (feature extraction); (2) analyze the electrocardiogram signals by extracting HRV metrics (feature extraction); (3) synchronize and fuse PA and HRV metrics (data fusion); (4) create a well-organized toolbox containing the developed algorithms for feature extraction and data fusion.

Skill required: Matlab programming, basic knowledge of biomechanics and signal processing. 


Physical behavior monitoring

No projects at the present time


Animal Tracking

No projects at the present time