Estimation of Walking Speed and Its Spatiotemporal Determinants Using a Single Inertial Sensor Worn on the Thigh: From Healthy to Hemiparetic Walking

Abstract

We present the use of a single inertial measurement unit (IMU) worn on the thigh to produce stride-by-stride estimates of walking speed and its spatiotemporal determinants (i.e., stride time and stride length). Ten healthy and eight post-stroke individuals completed a 6-min walk test with an 18-camera motion capture system used for ground truth measurements. Subject-specific estimation models were trained to estimate walking speed using the polar radius extracted from phase portraits produced from the IMU-measured thigh angular position and velocity. Consecutive flexion peaks in the thigh angular position data were used to define each stride and compute stride times. Stride-by-stride estimates of walking speed and stride time were then used to compute stride length. In both the healthy and post-stroke cohorts, low error and high consistency were observed for the IMU estimates of walking speed (MAE < 0.035 m/s; ICC > 0.98), stride time (MAE < 30 ms; ICC > 0.97), and stride length (MAE < 0.037 m; ICC > 0.96). This study advances the use of a single wearable sensor to accurately estimate walking speed and its spatiotemporal determinants during both healthy and hemiparetic walking.

Comparison of Three Motion Capture-Based Algorithms for Spatiotemporal Gait Characteristics: How Do Algorithms Affect Accuracy and Precision of Clinical Outcomes?

“…However, as mentioned in the introduction, MoCap-based algorithms have been used as a reference to validate other technologies or approaches in the literature [ 18 , 20 , 22 ]. As such, a spinoff of this study would be to define an early acceptable level of variation for new technology validation.…”

“…HS and TO represent initiation and termination instants of the foot contact to the ground, which corresponds to the change in detected force levels. Identification of HS and TO (i.e., the beginning of the rising edge and the end of the falling edge of the force signal, respectively) is often performed using a fixed (5 N [ 26 ], 10 N [ 27 , 28 , 29 ], 20 N [ 14 , 22 ], or adaptative percentage of the mass [ 30 ]) threshold. In the current study, these instants were detected using a plateau edge enhancement technique.…”

“…Thus, the scientific literature reveals that algorithms used for gait event detection perform differently, which in turn can impact the accuracy of the gait performance parameters extracted. Yet, several studies have used optoelectronic-based algorithms interchangeably either to validate another detection method or to answer a clinical question [ 15 , 18 , 19 , 20 , 21 , 22 ]. This is concerning as gait event detection algorithm robustness and accuracy appear to vary with population and gait velocity.…”

The effects of blurred visual inputs with different levels on the cerebral activity during free level walking

“…It had been reported that temporary blindfolding in individuals with normal vision would result in a decrease in the hip range of motion in the sagittal plane (Hallemans et al, 2010 ), which was consistent with our results. The hip was the proximal element in the kinematic chain of the lower extremities with significant importance in gait stability (Arumukhom Revi et al, 2021 ). The decreased range of motion in the hip indicated that, in the absence of visual navigation, an essential part of the motor adjustment was the limited movement in the supporting joint on the dynamic side to maintain stability.…”

Deep Learning Methods for Speed Estimation of Bipedal Motion from Wearable IMU Sensors

“…From a wide range of available methods, we analyze the performance of methods mentioned in survey [ 24 ], including extensive parameter tuning. Moreover, the default sensor placement in these approaches is almost always determined to be the foot attachment [ 4 ], with few exception such as [ 25 ]. We will study the sensitivity of this choice to estimation precision in the experimental section.…”