Human lower-extremity movement classification based on biomechanical sensor data: Machine learning approach

Abstract

Wearable biomechanical sensor signals can be used to precisely recognize human lower ex-tremity movements based upon gait parameters such as walking speed, which is an increasing-ly important field with a significant role in biomedical studies. In this study, human walking patterns were classified using wearable biomechanical sensors and machine learning and time series analysis techniques. Accurate classification of level-ground gait patterns of IMU, digital goniometer (GON) and electromyography (EMG) sensor data is of great importance in in-forming physicians and medical device innovators working in this discipline. For this study, an open access dataset recorded from four unilaterally placed IMUs, three GONs and eleven EMG sensors in 22 subjects at different walking speeds was used. The sliding time window method was used to extract features in the first part of biomedical signal processing. Then, the effects of various window lengths and single or multiple sensor models on machine learning classification performance are compared. The results of this study showed that the QSVM classifier and IMU-based sensor with a window length of 1000 (5s) had the highest classifi-cation accuracy of 0.954 to classify human gait at different walking speeds based on the pro-posed method. In addition, it is seen that the classifiers have different classification accuracy for the seven sensor models used. QSVM has higher accuracy in gait recognition compared to WNN and ESKNN classifiers. In particular, the accuracy (0.961) in the experiment using the IMU and GON multiple sensor and QSVM classifier is the highest among other sensor combinations and classifiers. When QSVM classification and gait recognition were compared, the accuracies were found as IMU (0.954), GON (0.827) and EMG (0.735) sensor models, respectively. Then, in dual sensor combination models, the highest accuracy was achieved in IMU-GON (0.961), IMU-EMG (0.895) and GON-EMG (0.776) sensor models, respectively. Finally, the accuracy of the IMU-GON-EMG model, in which all three sensors are included, is 0.919. The findings of this study showed that IMU sensor models improved the classification performance in level-ground gait pattern recognition, and their use together with GON sensor models contributed positively to this performance. It has been found that EMG sensor models show lower classification performance compared to IMU sensor modelsg the necessary pre-cautions were beneficial in terms of protecting the health of the employees.

Keywords

• Biomechanics
• Biomedical Signal Processing
• Human Movement
• Classification
• Machine Learning; Wearable Sensor

References

1. REFERENCES
2. [1] Thiyagarajan JA, Mikton C, Harwood RH, Gichu M, Gaigbe-Togbe V, Jhamba T, et al. The UN Decade of healthy ageing: Strengthening measurement for monitoring health and wellbeing of older people. Age Ageing 2022;51:afac147. [CrossRef]
3. [2] Chen X, Chen C, Wang Y, Yang B, Ma T, Leng Y, et al. A piecewise monotonic gait phase estimation model for controlling a powered transfemoral prosthesis in various locomotion modes. IEEE Robot Autom Lett 2022;7:95499456. [CrossRef]
4. [3] Boukhennoufa I, Zhai X, Utti V, Jackson J, McDonald-Maier KD. Wearable sensors and machine learning in post-stroke rehabilitation assessment: A systematic review. Biomed Signal Process Control 2022;71:103197. [CrossRef]
5. [4] Lempke AFD, Hart JM, Hryvniak DJ, Rodu JS, Hertel J. Use of wearable sensors to identify biomechanical alterations in runners with exercise-related lower leg pain. J Biomech 2021;126:110646. [CrossRef]
6. [5] Celik Y, Stuart S, Woo WL, Godfrey A. Gait analysis in neurological populations: Progression in the use of wearables. Med Eng Phys 2021;87:929.
7. [6] Teufl W, Lorenz M, Miezal M, Taetz B, Fröhlich M, Bleser G. Towards inertial sensor based mobile gait analysis: Event-detection and spatio-temporal parameters. Sensors (Basel) 2018;19:38. [CrossRef]
8. [7] Yang M, Zheng H, Wang H, McClean S, Hall J, Harris N. A machine learning approach to assessing gait patterns for complex regional pain syndrome. Med Eng Phys 2012;34:740746. [CrossRef] [8] Khan MH, Farid MS, Grzegorzek M. Spatiotemporal features of human motion for gait recognition. Signal Image Video Process 2019;13:369377. [CrossRef]

9. [9] Ferreira GA, Teixeira JLS, Rosso ALZ, de Sá AMF. On the classification of tremor signals into dyskinesia, Parkinsonian tremor, and Essential tremor by using machine learning techniques. Biomed Signal Process Control. 2022;73:103430. [CrossRef]
10. [10] Ma C, Li D, Pan L, Li X, Yin C, Li A, et al. Quantitative assessment of essential tremor based on machine learning methods using wearable device. Biomed Signal Process Control 2022;71:103244. [CrossRef]
11. [11] Nasrabadi AM, Eslaminia AR, Bakhshayesh PR, Ejtehadi M, Alibiglou L, Behzadipour S. A new scheme for the development of IMU-based activity recognition systems for telerehabilitation. Med Eng Phys 2022;108:103876. [CrossRef]
12. [12] Aliman N, Ramli R, Haris SM, Amiri MS, Van M. A robust adaptive-fuzzy-proportional-derivative controller for a rehabilitation lower limb exoskeleton. Eng Sci Technol Int J 2022;35:101097. [CrossRef]
13. [13] Nagaraj G, Mir BA, Gomathy B, Leelavathy S, Sengupta A, Ahmad SS. Artificial Neural Network to Predict Swinging of Lower Limb in Jumping Jack Exercise. In: 2023 International Conference on Intelligent Data Communication Technologies and Internet of Things (IDCIoT). IEEE; 2023. p. 914918. [CrossRef]
14. [14] Vijayvargiya A, Khimraj, Kumar R, Dey N. Voting-based 1D CNN model for human lower limb activity recognition using sEMG signal. Phys Eng Sci Med. 2021;44:12971309.
15. [15] Lencioni T, Carpinella I, Rabuffetti M, Marzegan A, Ferrarin M. Human kinematic, kinetic and EMG data during different walking and stair ascending and descending tasks. Sci Data 2019;6:309. [CrossRef]
16. [16] Dhiman C, Vishwakarma DK. A review of state-of-the-art techniques for abnormal human activity recognition. Eng Appl Artif Intell 2019;77:2145. [CrossRef]
17. [17] Kenan E, Kutlu MÇ, Barış B. Comparison of gesture classification methods with contact and non-contact sensors for human-computer interaction. Sigma J Eng Nat Sci 2022;40:219226. [CrossRef]
18. [18] Ma C, Li W, Cao J, Du J, Li Q, Gravina R. Adaptive sliding window based activity recognition for assisted livings. Inf Fusion 2020;53:5565. [CrossRef]
19. [19] Noor MHM, Salcic Z, Kevin I, Wang K. Adaptive sliding window segmentation for physical activity recognition using a single tri-axial accelerometer. Pervasive Mob Comput 2017;38:4159. [CrossRef]
20. [20] Wang G, Li Q, Wang L, Wang W, Wu M, Liu T. Impact of sliding window length in indoor human motion modes and pose pattern recognition based on smartphone sensors. Sensors (Basel) 2018;18:1965. [CrossRef]
21. [21] Huynh QT, Tran BQ. Time-frequency analysis of daily activities for fall detection. Signals 2021;2:112. [CrossRef] [22] Shawen N, O’Brien MK, Venkatesan S, Lonini L, Simuni T, Hamilton JL, et al. Role of data measurement characteristics in the accurate detection of Parkinson’s disease symptoms using wearable sensors. J Neuroeng Rehabil 2020;17:114. [CrossRef]
22. [23] Bhakta K, Camargo J, Compton W, Herrin K, Young A. Evaluation of continuous walking speed determination algorithms and embedded sensors for a powered knee & ankle prosthesis. IEEE Robot Autom Lett 2021;6:48204826. [CrossRef]
23. [24] Panyakaew P, Pornputtapong N, Bhidayasiri R. Using machine learning-based analytics of daily activities to identify modifiable risk factors for falling in Parkinson’s disease. Parkinsonism Relat Disord 2021;82:7783. [CrossRef]
24. [25] Trabassi D, Serrao M, Varrecchia T, Ranavolo A, Coppola G, De Icco R, et al. Machine learning approach to support the detection of Parkinson’s disease in IMU-based Gait analysis. Sensors (Basel) 2022;22:3700. [CrossRef]
25. [26] Xi X, Tang M, Miran SM, Luo Z. Evaluation of feature extraction and recognition for activity monitoring and fall detection based on wearable sEMG sensors. Sensors (Basel) 2017;17:1229. [CrossRef]
26. [27] Zhou B, Wang H, Hu F, Feng N, Xi H, Zhang Z, et al. Accurate recognition of lower limb ambulation mode based on surface electromyography and motion data using machine learning. Comput Methods Programs Biomed 2020;193:105486. [CrossRef]
27. [28] Camargo J, Flanagan W, Csomay-Shanklin N, Kanwar B, Young A. A machine learning strategy for locomotion classification and parameter estimation using fusion of wearable sensors. IEEE Trans Biomed Eng 2021;68:15691578. [CrossRef]
28. [29] Dong D, Ma C, Wang M, Vu HT, Vanderborght B, Sun Y. A low-cost framework for the recognition of human motion gait phases and patterns based on multi-source perception fusion. Eng Appl Artif Intell 2023;120:105886. [CrossRef] Camargo J, Ramanathan A, Flanagan W, Young A. A comprehensive, open-source dataset of lower limb biomechanics in multiple conditions of stairs, ramps, and level-ground ambulation and transitions. J Biomech 2021;119:110320. [CrossRef