Continuous Classification of Locomotion in Response to Task Complexity and Anticipatory State

Abstract

ObjectiveIntent recognition in lower-extremity assistive devices (e.g., prostheses and exoskeletons) is typically limited to either recognition of steady-state locomotion or changes of terrain (e.g., level ground to stair) occurring in a straight-line path and under anticipated condition. Stability is highly affected during non-steady changes of direction such as cuts especially when they are unanticipated, posing high risk of fall-related injuries. Here, we studied the influence of changes of direction and user anticipation on task recognition, and accordingly introduced classification schemes accommodating such effects.MethodsA linear discriminant analysis (LDA) classifier continuously classified straight-line walking, sidestep/crossover cuts (single transitions), and cuts-to-stair locomotion (mixed transitions) performed under varied task anticipatory conditions. Training paradigms with varying levels of anticipated/unanticipated exposures and analysis windows of size 100–600 ms were examined.ResultsMore accurate classification of anticipated relative to unanticipated tasks was observed. Including bouts of target task in the training data was necessary to improve generalization to unanticipated locomotion. Only up to two bouts of target task were sufficient to reduce errors to <20% in unanticipated mixed transitions, whereas, in single transitions and straight walking, substantial unanticipated information (i.e., five bouts) was necessary to achieve similar outcomes. Window size modifications did not have a significant influence on classification performance.ConclusionAdjusting the training paradigm helps to achieve classification schemes capable of adapting to changes of direction and task anticipatory state.SignificanceThe findings could provide insight into developing classification schemes that can adapt to changes of direction and user anticipation. They could inform intent recognition strategies for controlling lower-limb assistive to robustly handle “unknown” circumstances, and thus deliver increased level of reliability and safety.

The Effect of Sensor Feature Inputs on Joint Angle Prediction across Simple Movements

“…Including additional sensors may increase the computational complexity, which may increase the runtime for predicting user intent. Conversely, too few sensors may inadequately capture the user’s intent during complex tasks or task transitions [ 17 , 18 ].…”

“…Although the bidirectional LSTM resulted in a mean root mean squared error (RMSE) of 1.42–5.71°, the study was limited to level-ground walking and did not explore feature combinations to reduce the number of input signals [ 22 ]. Kazemimoghadam and Fey (2021) examined the effect of training task combinations for task recognition accuracy of transitions and found that including data of the target task in the training data enhanced the prediction accuracy on unanticipated locomotion tasks [ 17 ]. Mundt et al (2020) used an artificial neural network using IMU sensors to predict joint angles with a mean RMSE < 4.8° [ 23 ].…”

“…However, Dey et al (2021) did not examine the feature combinations of the input data, which could offer more granular-level information for HMIP. Kazemimoghadam and Fey (2021) observed signals of the lower limb and torso but did not compare the combinations of input signals to optimize prediction accuracy [ 17 ]. They did, however, recommend future studies to identify top input signals contributing to accurate prediction of the user’s intended tasks.…”