More Like this

1. Load Position and Weight Classification during Carrying Gait Using Wearable Inertial and Electromyographic Sensors

   https://doi.org/10.3390/s20174963  

   Goršič, Maja; Dai, Boyi; Novak, Domen (September 2020, Sensors)

   null (Ed.)

   Lifting and carrying heavy objects is a major aspect of physically intensive jobs. Wearable sensors have previously been used to classify different ways of picking up an object, but have seen only limited use for automatic classification of load position and weight while a person is walking and carrying an object. In this proof-of-concept study, we thus used wearable inertial and electromyographic sensors for offline classification of different load positions (frontal vs. unilateral vs. bilateral side loads) and weights during gait. Ten participants performed 19 different carrying trials each while wearing the sensors, and data from these trials were used to train and evaluate classification algorithms based on supervised machine learning. The algorithms differentiated between frontal and other loads (side/none) with an accuracy of 100%, between frontal vs. unilateral side load vs. bilateral side load with an accuracy of 96.1%, and between different load asymmetry levels with accuracies of 75–79%. While the study is limited by a lack of electromyographic sensors on the arms and a limited number of load positions/weights, it shows that wearable sensors can differentiate between different load positions and weights during gait with high accuracy. In the future, such approaches could be used to control assistive devices or for long-term worker monitoring in physically demanding occupations. 

   more » « less
2. Toward Real-Time Detection of Object Lifting Using Wearable Inertial Measurement Units

   https://doi.org/10.1109/EMBC46164.2021.9629585  

   Miller, Benjamin A.; Novak, Domen (November 2021, Proceedings of the 43rd Annual International Conference of the IEEE Engineering in Medicine and Biology Society)

   Back injuries and other occupational injuries are common in workers who engage in long, arduous physical labor. The risk of these injuries could be reduced using assistive devices that automatically detect an object lifting motion and support the user while they perform the lift; however, such devices must be able to detect the lifting motion as it occurs. We thus developed a system to detect the start and end of a lift (performed as a stoop or squat) in real time based on pelvic angle and the distance between the user's hands and the user's center of mass. The measurements were input to an algorithm that first searches for hand-center distance peaks in a sliding window, then checks the pelvic displacement angle to verify lift occurrence. The approach was tested with 5 participants, who performed a total of 100 lifts of four different types. The times of actual lifts were determined by manual video annotation. The median time error (absolute difference between detected and actual occurrence time) for lifts that were not false negatives was 0.11 s; a lift was considered a false negative if it was not detected within two seconds of it actually occurring. Furthermore, 95% of lifts that were detected occurred within 0.28 s of actual occurrence. This shows that it is possible to reliably detect lifts in real time based on the pelvic displacement angle and the distance between the user's hands and their center of mass. 

   more » « less
3. Wearable Sensor-Based Step Length Estimation During Overground Locomotion Using a Deep Convolutional Neural Network

   https://doi.org/10.1109/EMBC46164.2021.9630060  

   Jin, Heejoo; Kang, Inseung; Choi, Gayeon; Molinaro, Dean D.; Young, Aaron J. (November 2021, 2021 43rd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC))

   Step length is a critical gait parameter that allows a quantitative assessment of gait asymmetry. Gait asymmetry can lead to many potential health threats such as joint degeneration, difficult balance control, and gait inefficiency. Therefore, accurate step length estimation is essential to understand gait asymmetry and provide appropriate clinical interventions or gait training programs. The conventional method for step length measurement relies on using foot-mounted inertial measurement units (IMUs). However, this may not be suitable for real-world applications due to sensor signal drift and the potential obtrusiveness of using distal sensors. To overcome this challenge, we propose a deep convolutional neural network-based step length estimation using only proximal wearable sensors (hip goniometer, trunk IMU, and thigh IMU) capable of generalizing to various walking speeds. To evaluate this approach, we utilized treadmill data collected from sixteen able-bodied subjects at different walking speeds. We tested our optimized model on the overground walking data. Our CNN model estimated the step length with an average mean absolute error of 2.89 ± 0.89 cm across all subjects and walking speeds. Since wearable sensors and CNN models are easily deployable in real-time, our study findings can provide personalized real-time step length monitoring in wearable assistive devices and gait training programs. 

   more » « less
4. Modeling Fabric Movement for Future E-Textile Sensors

   https://doi.org/10.3390/s20133735  

   Ketola, Roope; Mishra, Vigyanshu; Kiourti, Asimina (July 2020, Sensors)

   null (Ed.)

   Studies with e-textile sensors embedded in garments are typically performed on static and controlled phantom models that do not reflect the dynamic nature of wearables. Instead, our objective was to understand the noise e-textile sensors would experience during real-world scenarios. Three types of sleeves, made of loose, tight, and stretchy fabrics, were applied to a phantom arm, and the corresponding fabric movement was measured in three dimensions using physical markers and image-processing software. Our results showed that the stretchy fabrics allowed for the most consistent and predictable clothing-movement (average displacement of up to −2.3 ± 0.1 cm), followed by tight fabrics (up to −4.7 ± 0.2 cm), and loose fabrics (up to −3.6 ± 1.0 cm). In addition, the results demonstrated better performance of higher elasticity (average displacement of up to −2.3 ± 0.1 cm) over lower elasticity (average displacement of up to −3.8 ± 0.3 cm) stretchy fabrics. For a case study with an e-textile sensor that relies on wearable loops to monitor joint flexion, our modeling indicated errors as high as 65.7° for stretchy fabric with higher elasticity. The results from this study can (a) help quantify errors of e-textile sensors operating “in-the-wild,” (b) inform decisions regarding the optimal type of clothing-material used, and (c) ultimately empower studies on noise calibration for diverse e-textile sensing applications. 

   more » « less
5. w-HAR: An Activity Recognition Dataset and Framework Using Low-Power Wearable Devices

   https://doi.org/10.3390/s20185356  

   Bhat, Ganapati; Tran, Nicholas; Shill, Holly; Ogras, Umit Y. (September 2020, Sensors)

   null (Ed.)

   Human activity recognition (HAR) is growing in popularity due to its wide-ranging applications in patient rehabilitation and movement disorders. HAR approaches typically start with collecting sensor data for the activities under consideration and then develop algorithms using the dataset. As such, the success of algorithms for HAR depends on the availability and quality of datasets. Most of the existing work on HAR uses data from inertial sensors on wearable devices or smartphones to design HAR algorithms. However, inertial sensors exhibit high noise that makes it difficult to segment the data and classify the activities. Furthermore, existing approaches typically do not make their data available publicly, which makes it difficult or impossible to obtain comparisons of HAR approaches. To address these issues, we present wearable HAR (w-HAR) which contains labeled data of seven activities from 22 users. Our dataset’s unique aspect is the integration of data from inertial and wearable stretch sensors, thus providing two modalities of activity information. The wearable stretch sensor data allows us to create variable-length segment data and ensure that each segment contains a single activity. We also provide a HAR framework to use w-HAR to classify the activities. To this end, we first perform a design space exploration to choose a neural network architecture for activity classification. Then, we use two online learning algorithms to adapt the classifier to users whose data are not included at design time. Experiments on the w-HAR dataset show that our framework achieves 95% accuracy while the online learning algorithms improve the accuracy by as much as 40%. 

   more » « less