| Full text | |
| Author(s): |
Perez-Ibarra, Juan C.
[1, 2]
;
Siqueira, Adriano A. G.
[2, 3, 4]
;
Krebs, Hermano I.
[1, 5, 6, 7, 8, 9, 10]
Total Authors: 3
|
| Affiliation: | [1] MIT, Dept Mech Engn, Cambridge, MA 02139 - USA
[2] Univ Sao Paulo, Dept Mech Engn, BR-13566590 Sao Carlos - Brazil
[3] Univ Sao Paulo, Ctr Adv Studies Rehabil, BR-13566590 Sao Carlos - Brazil
[4] Univ Sao Paulo, Ctr Robot Sao Carlos, BR-13566590 Sao Carlos - Brazil
[5] Univ Maryland, Sch Med, Dept Neurol, Baltimore, MD 21201 - USA
[6] Fujita Hlth Univ, Sch Med, Dept Rehabil Med 1, Toyoake, Aichi 4701192 - Japan
[7] Newcastle Univ, Inst Neurosci, Newcastle Upon Tyne NE1 7RU, Tyne & Wear - England
[8] Osaka Univ, Dept Mech Sci & Bioengn, Osaka 5650871 - Japan
[9] Wolfson Sch Mech Elect & Mfg, Loughborough LE11 3TU, Leics - England
[10] Sogang Univ, Coll Engn, Seoul 04107 - South Korea
Total Affiliations: 10
|
| Document type: | Journal article |
| Source: | IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING; v. 28, n. 12, p. 2933-2943, DEC 2020. |
| Web of Science Citations: | 0 |
| Abstract | |
Automatic identification of gait events is an essential component of the control scheme of assistive robotic devices. Many available techniques suffer limitations for real-time implementations and in guaranteeing high performances when identifying events in subjects with gait impairments. Machine learning algorithms offer a solution by enabling the training of different models to represent the gait patterns of different subjects. Here our aim is twofold: to remove the need for training stages using unsupervised learning, and to modify the parameters according to the changes within a walking trial using adaptive procedures. We developed two adaptive unsupervised algorithms for real-time detection of four gait events, using only signals from two single-IMU foot-mounted wearable devices. We evaluated the algorithms using data collected from five healthy adults and seven subjects with Parkinson's disease (PD) walking overground and on a treadmill. Both algorithms obtained high performance in terms of accuracy (F-1-score >= 0.95 for both groups), and timing agreement using a force-sensitive resistors as reference (mean absolute differences of 66 +/- 53 msec for the healthy group, and 58 +/- 63 msec for the PD group). The proposed algorithmsdemonstrated the potential to learn optimal parameters for a particular participant and for detecting gait eventswithout additional sensors, external labeling, or long training stages. (AU) | |
| FAPESP's process: | 15/50376-2 - Adaptive variable impedance applied to robotic rehabilitation of walking |
| Grantee: | Adriano Almeida Gonçalves Siqueira |
| Support Opportunities: | Regular Research Grants |
| FAPESP's process: | 13/14756-0 - Adaptive variable impedance applied to robotic rehabilitation of walking |
| Grantee: | Adriano Almeida Gonçalves Siqueira |
| Support Opportunities: | Regular Research Grants |