A Deep Learning Approach for Foot Trajectory Estimation in Gait Analysis Using Inertial Sensors

Guimar�ães, Vânia; Sousa, Inês; Correia, Miguel Velhote

doi:10.3390/s21227517

“…Higher scores of the metrics indicate better performances. To solve the network's overfitting problem and improve the generalization problem, the early stopping method [48] is used in the paper.…”

Section: Bat-inspired Elm Layersmentioning

confidence: 99%

Hybrid Optimized GRU-ECNN Models for Gait Recognition with Wearable IOT Devices

Monica

¹

,

Parvathi

²

,

Gayathri

³

et al. 2022

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

With the advent of the Internet of Things (IoT), human-assistive technologies in healthcare services have reached the peak of their application in terms of diagnosis and treatment process. These devices must be aware of human movements to provide better aid in clinical applications as well as the user’s daily activities. In this context, real-time gait analysis remains to be key catalyst for developing intelligent assistive devices. In addition to machine and deep learning algorithms, gait recognition systems have significantly improved in terms of high accuracy recognition. However, most of the existing models are focused on improving gait recognition while ignoring the computational overhead that affects the accuracy of detection and even remains unsuitable for real-time implementation. In this research paper, we proposed a hybrid gated recurrent unit (GRU) based on BAT-inspired extreme convolutional networks (BAT-ECN) for the effective recognition of human activities using gait data. The gait data are collected by implanting the wearable Internet of Things (WIoT) devices invasively. Then, a novel GRU and ECN networks are employed to extract the spatio-temporal features which are then used for classification to realize gait recognition. Extensive and comprehensive experimentations have been carried out to evaluate the proposed model using real-time datasets and also other benchmarks such as whuGait and OU-ISIR datasets. To prove the excellence of the proposed learning model, we have compared the model’s performance with the other existing hybrid models. Results demonstrate that the proposed model has outperformed the other learning models in terms of high gait classification and less computational overhead.

show abstract

“…Even without full understanding of the exact cause of the performance differences, it might be possible to develop algorithms that are less sensitive to changes in attachment. In particular, it would be interesting to see if data-driven methods based on machine learning, for example the algorithms presented in [ 36 – 38 ], are similarly sensitive to a change in sensor position as double-integration methods. Going beyond pure physics based calculation methods for spatial parameters might be a way to circumvent the described issues altogether, given sufficient training data for the individual attachment conditions.…”

Section: Resultsmentioning

confidence: 99%

The placement of foot-mounted IMU sensors does affect the accuracy of spatial parameters during regular walking

Küderle

¹

,

Roth

²

,

Zlatanovic

³

et al. 2022

View full text Add to dashboard Cite

Gait analysis using foot-worn inertial measurement units has proven to be a reliable tool to diagnose and monitor many neurological and musculoskeletal indications. However, only few studies have investigated the robustness of such systems to changes in the sensor attachment and no consensus for suitable sensor positions exists in the research community. Specifically for unsupervised real-world measurements, understanding how the reliability of the monitoring system changes when the sensor is attached differently is from high importance. In these scenarios, placement variations are expected because of user error or personal preferences. In this manuscript, we present the largest study to date comparing different sensor positions and attachments. We recorded 9000 strides with motion-capture reference from 14 healthy participants with six synchronized sensors attached at each foot. Spatial gait parameters were calculated using a double-integration method and compared to the reference system. The results indicate that relevant differences in the accuracy of the stride length exists between the sensor positions. While the average error over multiple strides is comparable, single stride errors and variability parameters differ greatly. We further present a physics model and an analysis of the raw sensor data to understand the origin of the observed differences. This analysis indicates that a variety of attachment parameters can influence the systems’ performance. While this is only the starting point to understand and mitigate these types of errors, we conclude that sensor systems and algorithms must be reevaluated when the sensor position or attachment changes.

show abstract

“…Even without full understanding of the exact cause of the performance differences, it might be possible to develop algorithms that are less sensitive to changes in attachment. In particular, it would be interesting to see if data-driven methods based on machine learning, for example the algorithms presented in [36][37][38], are similarly sensitive to a change in sensor position as double-integration methods. Going beyond pure physics based calculation methods for spatial parameters might be a way to circumvent the described issues altogether, given sufficient training data for the individual attachment conditions.…”

Section: Discussionmentioning

confidence: 99%

The placement of foot-mounted IMU sensors does affect the accuracy of spatial parameters during regular walking

Küderle

¹

,

Roth

²

,

Zlatanovic

³

et al. 2022

Preprint

View full text Add to dashboard Cite

Gait analysis using foot-worn inertial measurement units has proven to be a reliable tool to diagnose and monitor many neurological and musculoskeletal indications. However, only few studies have investigated the robustness of such systems to changes in the sensor attachment and no consensus for suitable sensor positions exists in the research community. Specifically for unsupervised real-world measurements, understanding how the reliability of the monitoring system changes when the sensor is attached differently is from high importance. In these scenarios, placement variations are expected because of user error or personal preferences. In this manuscript, we present the largest study to date comparing different sensor positions and attachments. We recorded 9000 strides with motion-capture reference from 14 healthy participants with six synchronized sensors attached at each foot. Spatial gait parameters were calculated using a double-integration method and compared to the reference system. The results indicate that relevant differences in the accuracy of the stride length exists between the sensor positions. While the average error over multiple strides is comparable, single stride errors and variability parameters differ greatly. We further present a physics model and an analysis of the raw sensor data to understand the origin of the observed differences. This analysis indicates that a variety of attachment parameters can influence the systems’ performance. While this is only the starting point to understand and mitigate these types of errors, we conclude that sensor systems and algorithms must be reevaluated when the sensor position or attachment changes.

show abstract

A Deep Learning Approach for Foot Trajectory Estimation in Gait Analysis Using Inertial Sensors

Cited by 18 publications

References 53 publications

Hybrid Optimized GRU-ECNN Models for Gait Recognition with Wearable IOT Devices

Hybrid Optimized GRU-ECNN Models for Gait Recognition with Wearable IOT Devices

The placement of foot-mounted IMU sensors does affect the accuracy of spatial parameters during regular walking

The placement of foot-mounted IMU sensors does affect the accuracy of spatial parameters during regular walking

Contact Info

Product

Resources

About