TY - GEN
T1 - Human Locomotion Assistance using Two-Dimensional Features Based Adaptive Oscillator
AU - Chinimilli, Prudhvi Tej
AU - Subramanian, Susheelkumar C.
AU - Redkar, Sangram
AU - Sugar, Thomas
N1 - Funding Information:
This work was partially supported by funding from DoD RIF program ”ExoSense: Sensing and Control Suite / Suit” (Contract No. W911NF-17-C-0044, primary contractor as NextGen Aeronautics and sub-contractor as Arizona State University). We acknowledge Dr. Jayanth Kudva, NextGen Aeronautics for his continuous support. The support from TPOCs Dr. Stephanie McElhinny (ARO), and CDR Jefferson Grubb (USSOCOM SOF ATL) is gratefully acknowledged.
Publisher Copyright:
© 2019 IEEE.
PY - 2019/5/21
Y1 - 2019/5/21
N2 - In this paper, an adaptive oscillator method Amplitude Omega Adaptive Oscillator (AωAO), is proposed to provide bilateral hip assistance for human locomotion. A realtime human locomotion recognition algorithm is integrated with AωAO to make it robust for various gait activities. The human locomotion recognition algorithm comprises both low-level (to detect activities) and high-level classifiers to detect transitions between activities. The Support Vector Machine (SVM) and Discrete Hidden Markov Model (DHMM) are used as low-level and high-level classifiers respectively. The human locomotion recognition algorithm is trained using two-dimensional features, Amplitude (A) and Omega (ω), obtained from thigh angle measurements, using a single Inertial Measurement Unit (IMU) on each limb. In AωAO, a pool with four adaptive oscillators (AOs) is used to estimate the filtered thigh angle trajectory. This pool converges to the frequency and phase of the signal, adaptively. To account for amplitude convergence, the amplitude parameters of the oscillator need to be reinitialized based on the human activity, identified by the human locomotion recognition algorithm. In addition to the adaptive oscillators, a Gaussian kernel function based nonlinear filter is employed to predict the future estimates of thigh angles. These predicted estimates, along with the user thigh angles, are used to calculate hip assistive torque in real-time. To verify the efficacy of the proposed approach, experiments were performed, using Hip exoskeleton for Superior Assistance (HeSA), on three healthy subjects. The human locomotion recognition algorithm reported higher classification and prediction accuracy of 95.2% and 94.9 % respectively. Activity Classification, Assistive devices, Human Activity Recognition.
AB - In this paper, an adaptive oscillator method Amplitude Omega Adaptive Oscillator (AωAO), is proposed to provide bilateral hip assistance for human locomotion. A realtime human locomotion recognition algorithm is integrated with AωAO to make it robust for various gait activities. The human locomotion recognition algorithm comprises both low-level (to detect activities) and high-level classifiers to detect transitions between activities. The Support Vector Machine (SVM) and Discrete Hidden Markov Model (DHMM) are used as low-level and high-level classifiers respectively. The human locomotion recognition algorithm is trained using two-dimensional features, Amplitude (A) and Omega (ω), obtained from thigh angle measurements, using a single Inertial Measurement Unit (IMU) on each limb. In AωAO, a pool with four adaptive oscillators (AOs) is used to estimate the filtered thigh angle trajectory. This pool converges to the frequency and phase of the signal, adaptively. To account for amplitude convergence, the amplitude parameters of the oscillator need to be reinitialized based on the human activity, identified by the human locomotion recognition algorithm. In addition to the adaptive oscillators, a Gaussian kernel function based nonlinear filter is employed to predict the future estimates of thigh angles. These predicted estimates, along with the user thigh angles, are used to calculate hip assistive torque in real-time. To verify the efficacy of the proposed approach, experiments were performed, using Hip exoskeleton for Superior Assistance (HeSA), on three healthy subjects. The human locomotion recognition algorithm reported higher classification and prediction accuracy of 95.2% and 94.9 % respectively. Activity Classification, Assistive devices, Human Activity Recognition.
KW - Activity Classification
KW - Assistive devices
KW - Human Activity Recognition
UR - http://www.scopus.com/inward/record.url?scp=85067106253&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85067106253&partnerID=8YFLogxK
U2 - 10.1109/WEARRACON.2019.8719628
DO - 10.1109/WEARRACON.2019.8719628
M3 - Conference contribution
AN - SCOPUS:85067106253
T3 - 2019 Wearable Robotics Association Conference, WearRAcon 2019
SP - 92
EP - 98
BT - 2019 Wearable Robotics Association Conference, WearRAcon 2019
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2019 Wearable Robotics Association Conference, WearRAcon 2019
Y2 - 25 March 2019 through 27 March 2019
ER -