lunduniversity.lu.se

Denna sida på svenska This page in English

Masters thesis projects

If you are interested in any of the projects below, please contact Christian Antfolk


Embedded signal processing of EMG signals using a Teensy 4.0 and ADS1299

Objective: Implement signal processing/machine learning/deep learning algorithms for advanced prosthesis control using an embedded platform

Approach: In this work, signal processing on the Arduino platform Teensy 4.0 will be performed. EMG-data will be captured from an ADS1299 chip from Texas Instruments and algorithms developed for the Teensy 4.0 microcontroller leveraging the CMSIS functionality of the M7 core . Implementation of time-domain and frequency domain features will be performed. Simple proportional control and machine learning algorithms will be performed on the platform.

Relevant skills: Electronics, signal processing, embedded software and programming experience with the Arduino environment and writing libraries. 

Equipment to be used: Teensy 4.0, ADS1299 and breadboard or similar prototyping equipment.

Relevant literature: ADS1299 datasheet, CMSIS manuals

E. Mastinu, B. Hakansson, and M. Ortiz-Catalan, “Low-cost, open source bioelectric signal acquisition system,” in 2017 IEEE 14th International Conference on Wearable and Implantable Body Sensor Networks (BSN), May 2017, pp. 19–22, doi: 10.1109/BSN.2017.7935997.

M. B. Pires, J. J. A. M. Junior, and S. L. Stevan, “Development of an 8 channel sEMG wireless device based on ADS1299 with Virtual Instrumentation,” Aug. 2018, Accessed: Oct. 20, 2021. [Online]. Available: arxiv.org/abs/1808.03711v1.

Contact information: Neuroengineering group, Christian Antfolk


Wireless EMG/EEG/ECG data-acquisition using the ADS1299

Objective: Develop a platform capable of performing data-acquisition and sending the acquired data wirelessly to a PC

Approach: An eight-channel EMG, ECG and EEG chip, the ADS1299, will be used in this work. An Arduino platform with Wifi/Bluetooth capabilities will be sourced in order to perform real-time data streaming from the platform to a PC. This work consists of evaluating different modes of communication (TCP/IP, UDP, data sockets etc) between the platform and the PC and validate the real-time streaming capability of the system.

Relevant skills and background: Communications, electronics and programming preferably experience with the Arduino environment and writing libraries.

Equipment to be used: Relevant Arduino type platform (Teensy, Feather etc) and ADS1299 chips, breadboards etc.

Relevant literature: Manuals, whitepapers

Z. Wang, W. Li, C. Chen, C. Sun, and W. Chen, “A multichannel reconfigurable EEG acquisition system design with felt-based soft material electrodes,” in 2018 IEEE International Instrumentation and Measurement Technology Conference (I2MTC), May 2018, pp. 1–6, doi: 10.1109/I2MTC.2018.8409883.

U. Rashid, I. Niazi, N. Signal, and D. Taylor, “An EEG Experimental Study Evaluating the Performance of Texas Instruments ADS1299,” Sensors, vol. 18, no. 11, p. 3721, Nov. 2018, doi: 10.3390/s18113721.

Contact information: Neuroengineering group, Christian Antfolk


Finger force regression using EMG + EEG

Objective: Record a dataset for finger force prediction based on EMG and EEG signals. (BSc) Develop an algorithm (or several algorithms) that could be used to detect onset of the movement, type (which finger is moving) and the force. (MSc)

Approach: Experimental work including the data-acquisition and signal processing of both EMG and EEG signals together with force exerted by the hand. A relevant experimental protocol needs to be defined based on experiment. Algorithms will be developed and tested in Python using relevant libraries (scikit-learn, numpy, scipy etc).

Relevant skills and background: Python programming, signal processing, aptitude for experimental work.

Equipment to be used: OtBioelettronica Quattrocento bioamplifier, force measuring device.

Relevant literature:
Malešević, N.; Andersson, G.; Björkman, A.; Controzzi, M.; Cipriani, C.; Antfolk, C. Instrumented Platform for Assessment of Isometric Hand Muscles Contractions. Meas. Sci. Technol. 2019, 30 (6), 065701. doi.org/10.1088/1361-6501/ab0eae.

OTBioelettronica Quattrocento Manual.

Contact information: Neuroengineering group, Christian Antfolk, Nebojsa Malesevic


Characterization of the artificial sensations produced by electrical stimulation 

Objective: Find optimal parameters (frequency, amplitude, electrode placement etc) of surface electrical stimulation in order to generate specific sensations.

Approach: This is exploratory work to find correlations between surface electrical stimulation parameters (such as amplitude, frequency, electrode position etc) and percepts of the stimulus. The main experimental equipment and GUI for executing tests are already developed and available. The project will rely heavily on experimental work, which will consist of tests on multiple volunteers and for different combinations of stimulation parameters. The protocol consists of placing one electrode over the median nerve, stimulating with a set of stimulation parameters, and obtaining feedback from the volunteer regarding the sensation that is artificially generated. Using the recorded database, a method should be developed to show influence of different stimulation parameters on sensation type and location.

Relevant skills and background: Programming skills (LabVIEW, Python) and knowledge of statistical methods

Equipment to be used: In house built stimulator, commercial stimulator, LabVIEW, Python

Relevant literature: Rudervall C. Can electrical pulses be adapted to create life-like perceptions? https://www.lunduniversity.lu.se/lup/publication/9033004

G. Chai, X. Sui, S. Li, L. He, and N. Lan, “Characterization of evoked tactile sensation in forearm amputees with transcutaneous electrical nerve stimulation,” J. Neural Eng., vol. 12, no. 6, 2015, doi: 10.1088/1741-2560/12/6/066002.

M. Johnson, “Transcutaneous Electrical Nerve Stimulation: Mechanisms, Clinical Application and Evidence,” Rev. Pain, vol. 1, no. 1, pp. 7–11, Aug. 2007, doi: 10.1177/204946370700100103.

Contact information: Neuroengineering group, Nebojsa Malesevic


Machine learning for multi-label classification of HD-sEMG signals with applications in prosthesis control

Background: Decoding movement intent from high-density surface EMG (HD-sEMG) can naturally be viewed as a multi-label classification problem, where each label represents a single kinematic degree of freedom.

Objective: Investigate, implement, and evaluate some selection of state-of-the-art multi-label or multi-output classification frameworks for the purpose of HD-sEMG movement decoding.

Approach: We have an in-house dataset of HD-sEMG with concurrent visual stimulus signals that will be used in this project. Algorithms to extract information pertaining to movement intent with respect to multiple degrees of freedom simultaneously from HD-sEMG will be developed, tuned, and tested. Development and testing will be carried out mainly in Python using appropriate software libraries (e.g. numpy/scipy, sklearn, keras, tensorflow, pytorch).

Relevant skills and background: Signal processing, Python programming, Statistics, Machine learning.

Equipment to be used: PC with relevant software packages (Python)

Relevant literature:
Olsson, A. E.; Sager, P.; Andersson, E.; Björkman, A.; Malešević, N.; Antfolk, C. Extraction of Multi-Labelled Movement Information from the Raw HD-sEMG Image with Time-Domain Depth. Sci. Rep. 2019, 9 7244. doi.org/10.1038/s41598-019-43676-8

Krasoulis, A.; Nazarpour, K. Myoelectric digit action decoding with multi-output, multi-class classification: an offline analysis. Sci. Rep. 2020, 10 16872. doi.org/10.1038/s41598-020-72574-7

Contact information: Neuroengineering group, Christian Antfolk, Alexander Olsson


Hand/wrist force estimation using intra-muscular EMG

Background: Intra-muscular electromyography (iEMG) can be used to measure latent muscle activity with high spatial resolution. It is possible to use iEMG to estimate the force with which a single forearm muscle contracts, although the degree to which activity in nearby muscles impact the resultant force exerted by the hand remains unknown.

Objective: Develop and evaluate learning algorithms for finger force regression based on iEMG.

Approach: We have an in-house dataset of forearm iEMG with concurrent finger force measurements that will be used in this project. Algorithms to infer the force generated based on iEMG will developed, tuned, and tested. Development and testing will be carried out mainly in Python using appropriate software libraries (e.g. numpy/scipy, sklearn, keras, tensorflow, pytorch).

Relevant skills and background: Signal processing, Python programming, Statistics, Machine learning.

Equipment to be used: PC with relevant software packages (MATLAB and Python)

Relevant literature:
Malešević, N.; Andersson, G.; Björkman, A.; Controzzi, M.; Cipriani, C.; Antfolk, C. Instrumented Platform for Assessment of Isometric Hand Muscles Contractions. Meas. Sci. Technol. 2019, 30 (6), 065701. doi.org/10.1088/1361-6501/ab0eae.

Malesevic, N.; Björkman, A.; Andersson, G. S.; Matran-Fernandez, A.; Citi, L.; Cipriani, C.; Antfolk, C. A Database of Multi-Channel Intramuscular Electromyogram Signals during Isometric Hand Muscles Contractions. Sci. Data 2020, 7 (1), 1–12. doi.org/10.1038/s41597-019-0335-8.

Olsson, A.; Malesevic, N.; Bjorkman, A.; Antfolk, C. Exploiting the Intertemporal Structure of the Upper-Limb SEMG: Comparisons between an LSTM Network and Cross-Sectional Myoelectric Pattern Recognition Methods. In Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS; IEEE, 2019; pp 6611–6615. doi.org/10.1109/EMBC.2019.8856648.

Contact information: Neuroengineering group, Christian Antfolk, Alexander Olsson


Fine-tuning of proportional myocontrol systems

Background: When using pattern recognition algorithms operating on myoelectric signals (EMG) for upper limb prosthesis control, data collection and model fitting are typically carried out only once, after which the resulting model is held constant. A real-time training phase, undertaken after the initial model fitting, could potentially be used to increase both model accuracy and the proficiency of the user.

Objective: Develop and evaluate an iterative recalibration procedure (with appertaining learning algorithms) for a myoelectric control interface.

Approach: An in-house dataset of surface EMG gathered with a Myo armband is available for offline evaluation of relevant myocontrol algorithms. Several algorithms to this end are already available, although modifications will have to be made in order to allow for online fine-tuning. Emphasis will be put on implementing appropriate countermeasures for known issues with online machine learning (e.g. data imbalance, catastrophic forgetting) relevant for the specific application. Development and testing will be carried out mainly in Python using appropriate software libraries (e.g. numpy/scipy, sklearn, keras, tensorflow, pytorch).

Relevant skills and background: Data acquisition, Signal processing, Experiment design, Python programming, Statistics, Machine learning.

Equipment to be used: PC with relevant software packages (Python), Thalmic Myo armband, Auxiliary biosignal acquisition systems (force rig, data glove, etc.).

Relevant literature:
Scheme, E.; Lock, B.; Hargrove, L.; Hill, W.; Kuruganti, U; Englehart, K. Motion Normalized Proportional Control for Improved Pattern Recognition-Based Myoelectric Control. IEEE Trans. Neural Syst. Rehabilitation Eng. 2013 22 (1). doi.org/10.1109/TNSRE.2013.2247421

Ameri, A.; Akhaee, A. A.; Scheme, E., Englehart, K. Regression convolutional neural network for improved simultaneous EMG control. J Neural Eng. 2019 16 (3). doi.org/10.1088/1741-2552/ab0e2e

Contact information: Neuroengineering group, Christian Antfolk, Alexander Olsson


Characterization of the hand grasping using sensor fusion

Objective: Record different hand grasps with multiple wearable sensors positioned on hand and wrist. Develop and evaluate algorithms that could be used to detect the type of grasp. 

Approach: Work on this project comprises familiarizing with different sensor technologies, including mm-wave radar, inertial measurement unit (IMU), and sensorized glove (Cyberglove). The technical part of the project includes (1) finding an optimal way of placing multiple sensors on hand, and (2) synchronizing sensors data during recording (or offline). The practical part of the project consists of the recording of different objects grasps by multiple volunteers. Using the recorded database, a machine learning based method should be developed to detect different grasp types.

Relevant skills and background: Programming skills (LabVIEW and/or Python, and/or Matlab), and machine learning skills

Equipment to be used: mm-wave radar senor (from Acconeer), IMU (from Bosch), and dataglove (Cyberglove)

Relevant literature: 
Y. Liu, L. Jiang, H. Liu, and D. Ming, “A Systematic Analysis of Hand Movement Functionality: Qualitative Classification and Quantitative Investigation of Hand Grasp Behavior,” Front. Neurorobot., vol. 15, p. 46, Jun. 2021, doi: 10.3389/fnbot.2021.658075.

F. Stival, S. Michieletto, M. Cognolato, E. Pagello, H. Müller, and M. Atzori, “A quantitative taxonomy of human hand grasps,” J. Neuroeng. Rehabil., vol. 16, no. 1, p. 28, Dec. 2019, doi: 10.1186/s12984-019-0488-x.

Contact information: Neuroengineering group, Nebojsa Malesevic


EMG control of prosthetic hand based on isotonic movements

Objective: Record a database of high-density surface EMG (HDsEMG) signals during isotonic contractions of different muscles. Implement and test existing algorithms to estimate muscle force and joint angle.

Approach: The first aim of this project is to record a database of HDsEMG signals during isotonic muscle contractions. The equipment for recording HDsEMG signals is available in the Neuroengineering laboratory, including EMG amplifier for 400 channels, high-density electrodes, recording software and weights. The following tasks should be conducted to record the database: (1) familiarization with the equipment and the software, (2) investigate which muscles can be recorded with the setup available at our laboratory, (3) defining of recording protocol, (4) recording of HDsEMG on at least 10 volunteers. The second aim of this project is to use the recorded database to design and test machine-learning algorithms for the estimation of load and joint angle. The basic algorithms will be given/provided by the supervisors. The task will be to tune the algorithms for this specific application and construct methods for training and testing of the algorithms (splitting the data, providing ground truth...).

Relevant skills and background: Programming skills (LabVIEW and/or Python, and/or Matlab), and machine learning skills

Equipment to be used: OtBioelettronica Quattrocento amplifier, joint angle measurement system (Cyberglove), PC with relevant software.

Relevant literature:

Contact information: Neuroengineering group, Christian Antfolk, Nebojsa Malesevic


Adaptation to perturbations during execution of functional tasks with prosthetic hand (while imposing perturbation of EMG control and feedback)

Objective: Conduct an experiment on the influence of perturbations in prosthetic control or feedback on the execution of functional tasks. Analyze the results of the experiment.

Approach: The aim of this project is to design, conduct and analyze the results of the experiment based on pick-and-place protocol. Healthy volunteers will be wearing a socket on their hand, on which a prosthetic hand will be mounted. By contracting their muscles, the hand will close and open. When squeezing a senzorized device, vibration feedback will be given to the user to indicate squeezing force. During the experiment, the control parameters (gains of EMG amplifiers that define transfer muscle contraction -> force of prosthetic hand) should be changed from time to time. Also, type feedback (number of feedback levels, direct or inverse proportionality between grip force and vibration strength) and will be changed from time to time. The goal is to evaluate how much time is needed for subjects to adapt to new control or feedback paradigm. To do so, several performance indicators will be tracked, such as duration of individual pick-and-place tasks, errors (dropped bottle). In addition, using eye-tracking device it will be possible to evaluate the amount of focus that a person dedicates to each task, indicating personal confidence in the control of prosthetic hand.

Relevant skills and background: Programming skills (Arduino, LabVIEW and/or Python, and/or Matlab)

Equipment to be used: Eye-tracking device (Tobii), Prosthetic hand (OttoBock), myocontrol sensors, miniature vibrational motor and sensorized object.

Relevant literature:
J. V. V. Parr, S. J. Vine, N. R. Harrison, and G. Wood, “Examining the Spatiotemporal Disruption to Gaze When Using a Myoelectric Prosthetic Hand,” J. Mot. Behav., vol. 50, no. 4, pp. 416–425, Jul. 2018, doi: 10.1080/00222895.2017.1363703.

Contact information: Neuroengineering group, Christian Antfolk, Nebojsa Malesevic


Eye-tracking for functional tests

Objective: Design a protocol for hand functional tests. Compare conventional performance metrics with the eye movement

Approach: The aim of the project is to find correlation (or absence of correlation) between conventional ways of quantifying performance during hand functional tests (such as number of repetitions in a specified time, number of errors...) and eye-movement parameters (such as number of saccades and fixations during each repetition). The recording protocol should be based on one of the common hand function test (such as ARAT).

Relevant skills and background: Signal processing (Python/Matlab), Statistics.

Equipment to be used: Eye-tracking device (Tobii or similar), objects and equipment required for performing functional tests.

Relevant literature:
Å. Nordin, M. A. Murphy, and A. Danielsson, “Intra-rater and inter-rater reliability at the item level of the Action Research arm test for patients with stroke,” J. Rehabil. Med., vol. 46, no. 8, pp. 738–745, Sep. 2014.

J. V. V. Parr, S. J. Vine, N. R. Harrison, and G. Wood, “Examining the Spatiotemporal Disruption to Gaze When Using a Myoelectric Prosthetic Hand,” J. Mot. Behav., vol. 50, no. 4, pp. 416–425, Jul. 2018, doi: 10.1080/00222895.2017.1363703.

Contact information: Neuroengineering group, Christian Antfolk, Nebojsa Malesevic


Generation of synthetic electromyography (EMG) data

Background: Generating reliable synthetic data is important for improving our understanding of real EMG recordings and will be directly relevant for mass testing of algorithms related to decomposition of such signals. The EMG signal is generated by motor units which are groups of cylindrical muscle fibers that contract in unison. The motor unit action potential is spread out both in time and along the length of the fibers and passes through the body which acts as a volume conductor. The implementation of a comprehensive and adaptable mathematical model from this thesis work, will be used to assess decomposition algorithms for large amounts of data with ground truth.

Objective: Implement a comprehensive and adaptable mathematical model for the generation of EMG data. Approach: The work will consist of defining and implementing a realistic mathematical model, with many variable parameters. This includes motor unit characteristics (like shape and firing frequency) and volume conductor parameters (like conductivity and tissue layers), as well as noise. The implementation will be done in MATLAB or Pyhon (numpy/scipy).

Relevant skills and background: Python/Matlab programming, (bio) signal processing and modelling.

Equipment to be used: PC with relevant software.

Relevant literature:
Merletti, R., Conte, L. L., Avignone, E., & Guglielminotti, P. (1999). Modeling of surface myoelectric signals. I. Model implementation. IEEE transactions on biomedical engineering, 46(7), 810-820.

Merletti, R., Roy, S. H., Kupa, E., Roatta, S., & Granata, A. (1999). Modeling of surface myoelectric signals. II. Model-based signal interpretation. IEEE Transactions on biomedical engineering, 46(7), 821-829.

Farina, D., Mesin, L., Martina, S., & Merletti, R. (2004). A surface EMG generation model with multilayer cylindrical description of the volume conductor. IEEE Transactions on Biomedical Engineering, 51(3), 415-426.

Contact information: Neuroengineering group, Christian Antfolk, Jonathan Lundsberg


Page Manager: Christian Antfolk