Detecting Toe-Off and Initial Contact in Real-Time With Self-Adapting Thresholds.

Foot angular kinematics measured with inertial measurement units: A reliable criterion for real-time gait event detection

Monitoring gait quality in daily activities through wearable sensors has the potential to improve medical assessment in Parkinson’s Disease (PD). In this study, four gait partitioning methods, two based on thresholds and two based on a machine learning approach, considering the four-phase model, were compared. The methods were tested on 26 PD patients, both in OFF and ON levodopa conditions, and 11 healthy subjects, during walking tasks. All subjects were equipped with inertial sensors placed on feet. Force resistive sensors were used to assess reference time sequence of gait phases. Goodness Index (G) was evaluated to assess accuracy in gait phases estimation. A novel synthetic index called Gait Phase Quality Index (GPQI) was proposed for gait quality assessment. Results revealed optimum performance (G < 0.25) for three tested methods and good performance (0.25 < G < 0.70) for one threshold method. The GPQI resulted significantly higher in PD patients than in healthy subjects, showing a moderate correlation with clinical scales score. Furthermore, in patients with severe gait impairment, GPQI was found higher in OFF than in ON state. Our results unveil the possibility of monitoring gait quality in PD through real-time gait partitioning based on wearable sensors.

Gait phase is an important state variable that helps exoskeletons provide assistance to users. Exoskeleton systems need to determine the assistance time and calculate the reference assistance force based on the gait phase. A large number of studies have used Finite State Machines (FSMs) or Adaptive Oscillators (AOs) to estimate gait phase. However, FSMs require manual calibration of gait events, and AOs can only estimate the phase of stable gait in real time. Both are difficult to adapt to irregular changes in gait. Deep learning algorithms can learn the nonlinear characteristics of kinematic data to estimate joint torque, gait phase, etc. The control algorithm based on Long Short-Term Memory (LSTM) Neural Network can automatically detect gait events according to the joint angle collected by Inertial Measurement Unit (IMU), effectively responding to dynamic changes in gait. This paper designed a Light-Weight Long Short-Term Memory (LWLSTM) Neural Network to estimate gait phase precisely and reduce training time of LSTM model. The experimental results showed that the Root Mean Square Error (RMSE) of LWLSTM was 8. 0 0 3 3%, quite lower than Convolutional Neural Network (CNN) and AO.

Real-time detection of gait events can be applied as a reliable input to control drop foot correction devices and lower-limb prostheses. Among the different sensors used to acquire the signals associated with walking for gait event detection, the accelerometer is considered as a preferable sensor due to its convenience of use, small size, low cost, reliability, and low power consumption. Based on the acceleration signals, different algorithms have been proposed to detect toe off (TO) and heel strike (HS) gait events in previous studies. While these algorithms could achieve a relatively reasonable performance in gait event detection, they suffer from limitations such as poor real-time performance and are less reliable in the cases of up stair and down stair terrains. In this study, a new algorithm is proposed to detect the gait events on three walking terrains in real-time based on the analysis of acceleration jerk signals with a time-frequency method to obtain gait parameters, and then the determination of the peaks of jerk signals using peak heuristics. The performance of the newly proposed algorithm was evaluated with eight healthy subjects when they were walking on level ground, up stairs, and down stairs. Our experimental results showed that the mean F1 scores of the proposed algorithm were above 0.98 for HS event detection and 0.95 for TO event detection on the three terrains. This indicates that the current algorithm would be robust and accurate for gait event detection on different terrains. Findings from the current study suggest that the proposed method may be a preferable option in some applications such as drop foot correction devices and leg prostheses.

Annotation of foot-contact and foot-off events is the initial step in post-processing for most quantitative gait analysis workflows. If clean force plate strikes are present, the events can be automatically detected. Otherwise, annotation of gait events is performed manually, since reliable automatic tools are not available. Automatic annotation methods have been proposed for normal gait, but are usually based on heuristics of the coordinates and velocities of motion capture markers placed on the feet. These heuristics do not generalize to pathological gait due to greater variability in kinematics and anatomy of patients, as well as the presence of assistive devices. In this paper, we use a data-driven approach to predict foot-contact and foot-off events from kinematic and marker time series in children with normal and pathological gait. Through analysis of 9092 gait cycle measurements we build a predictive model using Long Short-Term Memory (LSTM) artificial neural networks. The best-performing model identifies foot-contact and foot-off events with an average error of 10 and 13 milliseconds respectively, outperforming popular heuristic-based approaches. We conclude that the accuracy of our approach is sufficient for most clinical and research applications in the pediatric population. Moreover, the LSTM architecture enables real-time predictions, enabling applications for real-time control of active assistive devices, orthoses, or prostheses. We provide the model, usage examples, and the training code in an open-source package.

Gait analysis has traditionally been carried out in a laboratory environment using expensive equipment, but, recently, reliable, affordable, and wearable sensors have enabled integration into clinical applications as well as use during activities of daily living. Real-time gait analysis is key to the development of gait rehabilitation techniques and assistive devices such as neuroprostheses. This article presents a systematic review of wearable sensors and techniques used in real-time gait analysis, and their application to pathological gait. From four major scientific databases, we identified 1262 articles of which 113 were analyzed in full-text. We found that heel strike and toe off are the most sought-after gait events. Inertial measurement units (IMU) are the most widely used wearable sensors and the shank and foot are the preferred placements. Insole pressure sensors are the most common sensors for ground-truth validation for IMU-based gait detection. Rule-based techniques relying on threshold or peak detection are the most widely used gait detection method. The heterogeneity of evaluation criteria prevented quantitative performance comparison of all methods. Although most studies predicted that the proposed methods would work on pathological gait, less than one third were validated on such data. Clinical applications of gait detection algorithms were considered, and we recommend a combination of IMU and rule-based methods as an optimal solution.

The effectiveness of lower limb mobility assistance heavily relies on accurate gait pattern detection. Optimal support is achieved when sensors reliably detect gait information. This study aims to evaluate the accuracy of gait event detection using force plates compared to in-shoe IMUs under real-time conditions. Despite favorable sensor placement for real-time data collection, several challenges arose in assessing the system’s accuracy. Four healthy adults participated, performing self-paced walking on force plates while wearing in-shoe IMUs. Sagittal angular velocity (ω<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">sagittal</inf>) is the rate of rotational movement around the sagittal plane. As these measures demonstrated the most importance to accurate event identification, we concentrated on them in our investigation, together with the norm of jerk (J), as key characteristics for recognizing gait events using IMU data. We found that accurate gait event detection of initial contact and toe-off was achieved when the ω<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">sagittal</inf> was within the range of -300 to -340 [deg/sec] and the J fell between 100 and 440 [G/sec]. These specific parameter ranges were effective in providing precise gait event detection. The accuracy of these measurements was approximately 25 msec, with no misclassified events detected for the specified parameter ranges.

Recent advancements in computer vision led to the development of a real-time surveillance system which ensures the safety and security of the people in public places. An aerial surveillance system will be advantageous in this scenario using a platform like Unmanned Aerial Vehicle (UAV) will be very reliable and can be considered as a cost-effective option for this task. To make the system fully autonomous, we require real-time abnormal event detection. But, this is computationally complex and time-consuming due to the heavy load on the UAV, which affords limited processing and payload capacity. In this paper, we propose a cost-effective approach for aerial surveillance in which we move the large computation tasks to the cloud while keeping limited computation on-board UAV device using edge computing technique. Further, our proposed system will maintain the minimum communication between UAV and cloud. Thus it not only reduces the network traffic but also reduces the end-to-end delay. The proposed method is based on the state-of-the-art YOLO (You Only Look Once) technique for real-time object detection deployed on edge computing device using Intel neural compute stick Movidius VPU (Vision Processing Unit), and we applied abnormal event detection using motion influence map on the cloud. Experimental results demonstrate that the proposed system reduces the end-to-end delay. Further, Tiny YOLO is six times faster while processing the frames per second (fps) when compared to other state-of-the-art methods.

Gait rehabilitation commonly relies on bodyweight unloading mechanisms, such as overhead mechanical support and underwater buoyancy. Lightweight and wireless inertial measurement unit (IMU) sensors provide a cost-effective tool for quantifying body segment motions without the need for video recordings or ground reaction force measures. Identifying the instant when the foot contacts and leaves the ground from IMU data can be challenging, often requiring scrupulous parameter selection and researcher supervision. We aimed to assess the use of machine learning methods for gait event detection based on features from foot segment rotational velocity using foot-worn IMU sensors during bodyweight-supported treadmill walking on land and underwater. Twelve healthy subjects completed on-land treadmill walking with overhead mechanical bodyweight support, and three subjects completed underwater treadmill walking. We placed IMU sensors on the foot and recorded motion capture and ground reaction force data on land and recorded IMU sensor data from wireless foot pressure insoles underwater. To detect gait events based on IMU data features, we used random forest machine learning classification. We achieved high gait event detection accuracy (95–96%) during on-land bodyweight-supported treadmill walking across a range of gait speeds and bodyweight support levels. Due to biomechanical changes during underwater treadmill walking compared to on land, accurate underwater gait event detection required specific underwater training data. Using single-axis IMU data and machine learning classification, we were able to effectively identify gait events during bodyweight-supported treadmill walking on land and underwater. Robust and automated gait event detection methods can enable advances in gait rehabilitation.

The purpose of this study was to develop and validate an automatic procedure using accelerometers and angular rate sensors for detecting exact gait events: heel contact (HC), toe contact (TC), heel off (HO) and toe off (TO). For this purpose, two miniature-sized light sensors were attached to the right shoe which measured triaxial acceleration and biaxial angular velocity of various positions of the shoe while ground reaction force and marker data were recorded at the same time as reference measurements. Healthy men walked on a walkpath at self-selected normal speed with the instrumented shoes on. An automated algorithm was developed to detect HC, TC, HO and TO from the acceleration and angular velocity signals, which were compared with reference data from force plates and markers. The results showed that our method demonstrated similar HC and TO to those with force plates and markers. Moreover, this method well detected stance phase duration. However, TC and HO differed among the three methods (force plates vs. markers vs. sensors). We concluded that the system using inertia sensors can identify TC and HO accurately and be used in gait analysis.

Accurate detection of human motion states and prediction of intentions are crucial for the natural walking assistance provided by lower-limb exoskeletons. Several methods for gait task recognition, phase estimation, and motion trajectory prediction have been proposed, but few studies have considered all of these simultaneously. This article proposes a novel scheme for gait recognition and trajectory prediction, capable of detecting gait events, recognizing multiple gait tasks, estimating gait phases, and predicting trajectories. First, this scheme proposes a gait task recognition method based on gait events and a multilayer perceptron neural network (MLPNN), which reduces computational cost and improves recognition stability. Next, a gait phase estimation method based on resettable symmetric adaptive oscillators (RSAOs) is introduced, which offers faster convergence speed and smaller error compared to traditional adaptive oscillators (AOs). Subsequently, a trajectory prediction method with adaptive dynamic motion primitives (ADMP) and feedback regulation is proposed. Finally, gait task recognition, phase estimation, trajectory prediction, and comprehensive online experiments are carried out respectively. The online experimental results show that the recognition accuracy of gait tasks can reach 99.2%, the phase estimation error is less than 0.3652 rad, and the trajectory estimation error is less than 1.1814°.

Examination of inertial measurement units to evaluate lower body segmental angles in persons with autism.

Although real-time event detection during video EEG recording is required to ensure patients’ safety, it is limited by the technologists’ availability. We sought to explore the efficiency of real-time event detection by the EEG technologists in a single tertiary academic center. We retrospectively reviewed events from continuous inpatient video EEGs (cEEGs) and epilepsy monitoring unit (EMU) recordings in January 2017, when real-time surveillance was only available during the night shift, and June 2017, when a dedicated neurodiagnostic EEG technologist was available for real-time monitoring during all shifts. The events were categorized into those detected immediately (eyes-on), later in the same shift (delayed) or identified on the subsequent shift (missed). Chi-square and Fisher’s exact tests were used for statistical comparisons. In January 2017, there were 25 patients (117 days of monitoring) in the EMU and 54 inpatients (146 days of monitoring) on cEEG with 92 total events, (39% seizures). In June 2017, there were 30 patients (133 days of monitoring) in the EMU and 47 additional inpatients (80 days of monitoring) on cEEG with 110 total events, (39% seizures). The number of events identified in real time was low and did not significantly differ among shifts regardless of the availability of the monitoring technologist. Most events were identified at the time of subsequent EEG scanning by the EEG technologist. Partial staffing for continuous video EEG surveillance is insufficient to identify events in real time. EEG technologists are able to identify events during regular EEG scanning.

Adaptive gait detection based on foot-mounted inertial sensors and multi-sensor fusion

Wearable sensors are able to monitor physical health in a home environment and detect changes in gait patterns over time. To ensure long-term user engagement, wearable sensors need to be seamlessly integrated into the user's daily life, such as hearing aids or earbuds. Therefore, we present EarGait, an open-source Python toolbox for gait analysis using inertial sensors integrated into hearing aids. This work contributes a validation for gait event detection algorithms and the estimation of temporal parameters using ear-worn sensors. We perform a comparative analysis of two algorithms based on acceleration data and propose a modified version of one of the algorithms. We conducted a study with healthy young and elderly participants to record walking data using the hearing aid's integrated sensors and an optical motion capture system as a reference. All algorithms were able to detect gait events (initial and terminal contacts), and the improved algorithm performed best, detecting 99.8% of initial contacts and obtaining a mean stride time error of 12 ± 32 ms. The existing algorithms faced challenges in determining the laterality of gait events. To address this limitation, we propose modifications that enhance the determination of the step laterality (ipsi- or contralateral), resulting in a 50% reduction in stride time error. Moreover, the improved version is shown to be robust to different study populations and sampling frequencies but is sensitive to walking speed. This work establishes a solid foundation for a comprehensive gait analysis system integrated into hearing aids that will facilitate continuous and long-term home monitoring.