Tesi etd-09272021-152140
Link copiato negli appunti
Tipo di tesi
Dottorato
Autore
TRUPPA, LUIGI
URN
etd-09272021-152140
Titolo
Development of Innovative Inertial Sensors methods to Monitor Human Motion and Athletic Movement
Settore scientifico disciplinare
ING-IND/34
Corso di studi
Istituto di Biorobotica - BIOROBOTICS
Commissione
relatore Dott. MANNINI, ANDREA
Parole chiave
- Nessuna parola chiave trovata
Data inizio appello
10/01/2022;
Disponibilità
parziale
Riassunto analitico
This doctoral thesis focus on the application of Inertial Measurement Units (IMUs) in sports biomechanics. The basic hypothesis behind the work is that it is possible to monitor the performances of athletes by using wearable sensors. In particular, during this three year PhD program I focused on the kinematic analysis of the athletic gesture. After the scientific literature analysis, the choice felt on Magneto-Inertial Measurement Units (MIMUs) based on micro-electro-mechanical (MEMS) technology since they result very versatile, cheap and small in size sensors. Despite wearables are characterized by a lower accuracy compared to optoelectronic methods, which can be considered as the gold standard for kinematic assessment, they can be used in uncontrolled environments, enabling the possibility to test methods directly on the playground . Thus, they require a particular care about computational strategies to extract reliable information from them. In particular, MIMUs rely on magnetometers which are strongly affected by magnetic disturbances. Indeed, the presence of ferromagnetic objects could alter considerably the Earth magnetic field distribution, causing the erroneous definition of the MIMU local reference frame. This phenomenon greatly limits the indoor applications of MIMU technology, especially in sport scenarios, in which the presence of ferromagnetic objects (e.g., sport equipment) is strong. That’s the reason why I focused my attention on an alternative technology, namely Inertial Measurement Units (IMUs), which rely only on gyroscopes and accelerometers neglecting magnetometers contribution. Unfortunately, IMUs are still affected by several sources of errors, such as bad calibration.
In order test the reliability of IMUs for human kinematic assessment, I developed a magnetic-free version of the most used Kalman filters (i.e., Extended and Unscented Kalman Filters) complementarily with a specific biomechanical model used to align sensors sensing axes to anatomical frames, based on simple functional movements. In particular, methodological solutions were tested on the upper limb of several participants performing yoga sequences. This choice was motivated by the need to test the algorithms during a complex movement characterized by a sequence of different body postures, maintained over time, and characterized by rotations in all three directions of the 3D space. The correctness of the estimations was evaluated by comparing the joint angles measured using IMUs with the ones obtained from an opto-electronic system (OMC), considered as gold standard (i.e., evaluating Mean Absolute Error and Pearson’s correlation coefficient between methods). The low values of Mean Absolute Error (< 10°) and the high Pearson’s correlation coefficient (> 0.86) validated the Extended (EKF) and Unscented (UKF) Kalman filters as a suitable solution for human kinematic analysis, especially considering the high range of motion of the proposed yoga sequences (up to 200°). Once verified that a magnetic-free solution for human motion capture is possible, I developed a new innovative sensor fusion algorithms for kinematic assessment, which embodies also an online module for bias detection and compensation. The proposed algorithm fully succeed in estimating the bias under both static and dynamic conditions, allowing a better extrapolation of the upper limb joint angles in the same yoga sequences. Here again, the low values of the Mean Absolute Error (< 7°) and the high Pearson’s correlation coefficient (> 0.91) allowed us to validate the proposed sensor fusion algorithm as a suitable methodology for kinematic assessment. The kinematic analysis of the upper limb joint angles was also extended to archery, which has several similarities with yoga. Error analysis (i.e., Mean absolute error and Pearson’s correlation coefficient) between IMUs and OMC confirmed the inertial body sensor network as a suitable technology for kinematic capture in competitive sports, such as archery.
To further extend the work, the second part of my Ph.D. thesis involved the application of the proposed methodologies for kinematic assessment to the practical problem of fatigue detection/quantification in more complex sports (i.e., soccer). As concerns fatigue detection, the performance in vertical jump in sixteen soccer referees were monitored before and after a soccer-specific fatigue protocol by using a body sensor network of six IMUs placed on the participants’ lower limbs. The study of the IMUs outputs (i.e., accelerations, angular velocities and joint angles) allowed us to define nineteen instrumental “markers” for fatigue detection. The analysis of their variations allowed us to distinguish two opposites but coherent responses to the fatigue protocol. Indeed, eight out of sixteen athletes showed reduced performance (e.g., an effective fatigue condition), while the other eight athletes experienced an improvement of the execution likely due to the so-called Post-Activation Potentiation. In both cases, the above parameters were significantly influenced by the fatigue protocol (p < 0.05), confirming their validity for fatigue monitoring. Interesting correlations between several biomechanical parameters and muscular mass were highlighted in the fatigued group. Finally, a “fatigue approximation index” was proposed and validated as fatigue quantifier. Another study I propose in this thesis aimed to give a first instrumental analysis of the physiological and biomechanical adaptation of football players to a fatigue protocol during the month immediately after the COVID-19 lockdown, to get insights about the fitness recovery. Eight male semi-professional football players took part in the study and filled a questionnaire about their activity during the lockdown. At activities resume, the mean heart rate and covered distances during fatiguing exercises, the normalized variations of mean and maximum exerted power in the Wingate test and the Bosco test out-comes (i.e., maximum height, mean exerted power, relative strength index, leg stiffness, contact time and flight time) were measured for one month. A significant effect of fatigue (Wilcoxon signed rank test p<0.05) on measured variables was confirmed for the four weeks. The analysis of the normalized variations of the aforementioned parameters allowed distinguishing two behaviours: downfall in the first two weeks and recovery in the last two weeks. Instrumental results showed a physiological and ballistic (i.e., Bosco test outcomes) recovery after four weeks.
In the end, my Ph.D. work allowed to design and develop new innovative algorithms for IMUs application in sports biomechanics and, so, in indoor and uncontrolled environments. The proposed methods have proven to be reliable and effective in real-world scenarios (such as yoga, soccer and archery), validating an inertial body sensor network for kinematic assessment and performance monitoring in different kind of athletes or home-users.
In order test the reliability of IMUs for human kinematic assessment, I developed a magnetic-free version of the most used Kalman filters (i.e., Extended and Unscented Kalman Filters) complementarily with a specific biomechanical model used to align sensors sensing axes to anatomical frames, based on simple functional movements. In particular, methodological solutions were tested on the upper limb of several participants performing yoga sequences. This choice was motivated by the need to test the algorithms during a complex movement characterized by a sequence of different body postures, maintained over time, and characterized by rotations in all three directions of the 3D space. The correctness of the estimations was evaluated by comparing the joint angles measured using IMUs with the ones obtained from an opto-electronic system (OMC), considered as gold standard (i.e., evaluating Mean Absolute Error and Pearson’s correlation coefficient between methods). The low values of Mean Absolute Error (< 10°) and the high Pearson’s correlation coefficient (> 0.86) validated the Extended (EKF) and Unscented (UKF) Kalman filters as a suitable solution for human kinematic analysis, especially considering the high range of motion of the proposed yoga sequences (up to 200°). Once verified that a magnetic-free solution for human motion capture is possible, I developed a new innovative sensor fusion algorithms for kinematic assessment, which embodies also an online module for bias detection and compensation. The proposed algorithm fully succeed in estimating the bias under both static and dynamic conditions, allowing a better extrapolation of the upper limb joint angles in the same yoga sequences. Here again, the low values of the Mean Absolute Error (< 7°) and the high Pearson’s correlation coefficient (> 0.91) allowed us to validate the proposed sensor fusion algorithm as a suitable methodology for kinematic assessment. The kinematic analysis of the upper limb joint angles was also extended to archery, which has several similarities with yoga. Error analysis (i.e., Mean absolute error and Pearson’s correlation coefficient) between IMUs and OMC confirmed the inertial body sensor network as a suitable technology for kinematic capture in competitive sports, such as archery.
To further extend the work, the second part of my Ph.D. thesis involved the application of the proposed methodologies for kinematic assessment to the practical problem of fatigue detection/quantification in more complex sports (i.e., soccer). As concerns fatigue detection, the performance in vertical jump in sixteen soccer referees were monitored before and after a soccer-specific fatigue protocol by using a body sensor network of six IMUs placed on the participants’ lower limbs. The study of the IMUs outputs (i.e., accelerations, angular velocities and joint angles) allowed us to define nineteen instrumental “markers” for fatigue detection. The analysis of their variations allowed us to distinguish two opposites but coherent responses to the fatigue protocol. Indeed, eight out of sixteen athletes showed reduced performance (e.g., an effective fatigue condition), while the other eight athletes experienced an improvement of the execution likely due to the so-called Post-Activation Potentiation. In both cases, the above parameters were significantly influenced by the fatigue protocol (p < 0.05), confirming their validity for fatigue monitoring. Interesting correlations between several biomechanical parameters and muscular mass were highlighted in the fatigued group. Finally, a “fatigue approximation index” was proposed and validated as fatigue quantifier. Another study I propose in this thesis aimed to give a first instrumental analysis of the physiological and biomechanical adaptation of football players to a fatigue protocol during the month immediately after the COVID-19 lockdown, to get insights about the fitness recovery. Eight male semi-professional football players took part in the study and filled a questionnaire about their activity during the lockdown. At activities resume, the mean heart rate and covered distances during fatiguing exercises, the normalized variations of mean and maximum exerted power in the Wingate test and the Bosco test out-comes (i.e., maximum height, mean exerted power, relative strength index, leg stiffness, contact time and flight time) were measured for one month. A significant effect of fatigue (Wilcoxon signed rank test p<0.05) on measured variables was confirmed for the four weeks. The analysis of the normalized variations of the aforementioned parameters allowed distinguishing two behaviours: downfall in the first two weeks and recovery in the last two weeks. Instrumental results showed a physiological and ballistic (i.e., Bosco test outcomes) recovery after four weeks.
In the end, my Ph.D. work allowed to design and develop new innovative algorithms for IMUs application in sports biomechanics and, so, in indoor and uncontrolled environments. The proposed methods have proven to be reliable and effective in real-world scenarios (such as yoga, soccer and archery), validating an inertial body sensor network for kinematic assessment and performance monitoring in different kind of athletes or home-users.
File
| Nome file | Dimensione |
|---|---|
Ci sono 1 file riservati su richiesta dell'autore. |
|