Identification Algorithm for Walking Environments : A Statistical and Artificial Intelligent Approach
- 발행기관 서강대학교 일반대학원
- 지도교수 신충수
- 발행년도 2024
- 학위수여년월 2024. 8
- 학위명 박사
- 학과 및 전공 일반대학원 기계공학과
- 실제 URI http://www.dcollection.net/handler/sogang/000000078955
- UCI I804:11029-000000078955
- 본문언어 영어
- 저작권 서강대학교 논문은 저작권 보호를 받습니다.
초록 (요약문)
Walking is not only the most fundamental activity ensuring mobility in daily life but also one of the primary means for mobility. However, walking ability decreases due to muscle weakness or injury in the lower extremity. The walking assistive device enhances a user's walking ability and physical capabilities by providing additional torque to the lower limb joints, thereby allowing the user to walk safely. The walking assistive device user inevitably encounters various walking environments. Since there are kinematic and kinetic differences in lower extremity joints according to various walking environments, walking assistive devices cannot accomplish the desired control strategy when they do not obtain information about walking environments properly. Thus, the objective of this dissertation was to propose an identification algorithm for classifying current environments and transitions between various environments during walking. Firstly, pressure plate was used to obtain the plantar pressure while the participants walked on flat ground and transition from flat ground to downstairs. Based on the plantar pressure, the stance time, vertical ground reaction force, anteroposterior (AP) and mediolateral (ML) center of pressure at initial contact, and AP/ML range of COP were calculated for explanatory variables. After analyzing the results of boxplot and paired t-test for the explanatory variables, the stance time, the ML and AP COP at IC, and AP range of COP were significantly different between LW and TW-LSD (all, p<0.001). Logistic regression models were generated using response variables and explanatory variables. Accuracy, sensitivity, and specificity were 98.2%, 100%, and 96.4%, respectively, as the result of walking condition classification using this model. Thus, the model generated by the logistic regression method could distinguish between the LW and TW-LSD with high accuracy (specific aim 1). Secondly, the muscle activations of selected muscles in the lower extremities (rectus femoris, vastus medialis and lateralis, semitendinosus, biceps femoris, tibialis anterior, soleus, medial and lateral gastrocnemius, flexor hallucis longus, and extensor digitorum longus) were measured in 27 participants while they walked over flat-ground, upstairs, downstairs, uphill, and downhill. An artificial neural network (ANN) was employed to classify these walking environments using the entire surface electromyography (sEMG) profile recorded for all muscles during the stance phase. The result shows that the ANN was able to classify the current walking environment with high accuracy of 96.3% when using activation from all muscles. When muscle activation from flexor/extensor groups in the knee, ankle, and metatarsophalangeal joints were used individually to classify the environment, the triceps surae muscle activation showed the highest classification accuracy of 88.9%. Thus, a current walking environment was classified with high accuracy using an ANN based on only sEMG signals (specific aim 2). Lastly, the muscle activations of selected muscles in the lower extremities (rectus femoris, vastus medialis and lateralis, semitendinosus, biceps femoris, tibialis anterior, soleus, medial and lateral gastrocnemius, flexor hallucis longus, and extensor digitorum longus) of 27 subjects were measured while walking on flat ground, upstairs, downstairs, uphill, and downhill and transitioning between these walking surfaces. An artificial neural network (ANN) was used to construct the model, taking the entire sEMG profile during the stance phase as input, to identify transitions between walking environments. The results show that transitioning between walking environments, including continuously walking on a current terrain, was successfully classified with high accuracy of 95.4 % when using all muscle activations. When using a combination of muscle activations of the knee extensor, ankle extensor, and metatarsophalangeal flexor group as classifying parameters, the classification accuracy was 90.9 %. Thus, transitioning between gait environments could be identified with high accuracy with the ANN model using only sEMG signals measured during the stance phase (specific aim 3). As a whole, the algorithms developed using both statistical and artificial intelligence approaches in this study were effective in distinguishing various walking environments. It is considered that the approach based on this study could be taken into account for controlling walking assistive devices to ensure smooth movement in response to changing walking environments. Based on the EMG signals and ANN proposed in this study, it is possible to build a real-time identification algorithm for the walking environment. It is also believed to help control walking assistance devices in real-time according to changes in the walking environment.
more목차
I. Chapter 1 Introduction 1
1.1 Statement of the problems 1
1.2 Specific Aims 3
1.2.1 Specific Aim1 3
1.2.2 Specific Aim2 4
1.2.3 Specific Aim3 5
II. Chapter 2 Literature Review 7
2.1 Detection algorithm that distinguishes between flat walking and transition walking from level to stair descent (specific aim1) 7
2.2 Classification algorithm for the current walking environment (specific aim2) 9
2.3 Identification algorithm for transitions between various walking environments (specific aim3) 12
III. Chapter 3 Classification of Level Walking and Transition Walking from Level to Stair Descent based on Plantar Pressure (specific aim1) 15
3.1 Materials and Methods 16
3.1.1 Participants 16
3.1.2 Experimental Protocol 17
3.1.3 Calculation of Parameter using Plantar Pressure and Generation of Variables 19
3.1.3.1 Definition of the origin of the foot local coordinate system 19
3.1.3.2 Calculation of Vertical Ground Reaction Force 19
3.1.3.3 Calculation of Center of Pressure and Range of Center of Pressure 20
3.1.3.4 Generation of Variables 22
3.1.4 Statistical Analysis and Model Building 23
3.1.4.1 Statistical Analysis 23
3.1.4.2 Model building 25
3.2 Results 27
3.3 Discussion 28
IV. Chapter 4 Classification of Walking Environments Using Deep Learning Approach Based on Surface EMG Sensors Only (specific aim2) 31
4.1 Materials and Methods 32
4.1.1 Participants 32
4.1.2 Experimental Protocol 32
4.1.3 Data Collection 34
4.1.4 Data Processing 35
4.1.5 Walking Environment Classification 38
4.2 Results 43
4.3 Discussion 47
V. Chapter 5 Deep Learning-based Identification Algorithm for Transitions between Walking Environments using Electromyography Signals only (specific aim3) 51
5.1 Materials and Methods 52
5.1.1 Participants 52
5.1.2 Experimental Protocol 52
5.1.3 Data Collection 55
5.1.4 Data Processing 55
5.1.5 Identification of Walking Tasks 56
5.2 Results 61
5.3 Discussion 66
VI. Chapter 6 Conclusion 71
VII. Chapter 7 Applications 74
7.1 Possibility of the Real-time Classification 74
7.2 Muscle fatigue monitoring system 75
References 76
