Gait Segmentation Research Articles

Machine Learning based Human Gait Segmentation with Wearable Sensor Platform.

Supervised and unsupervised machine learning algorithms were explored for gait segmentation using wearable sensor platform. Multiple wearable sensors modules were placed at key locations: Four Inertial Measurement Units (IMUs) were attached to the thigh and shank of each leg and a plantar pressure measuring foot insoles were implanted in the shoes. The gait data has been collected from 10 people wirelessly via TCI-IP protocol, which is later anonymized. Further, the Ranchos Los Amigos (RLA) gait nomenclature-based data preprocessing and peak/valley detector based annotation steps are performed on the acquired data followed by implementation of machine learning techniques on the labeled datasets. The methods explored for phase and sub-phase classification includes the Unsupervised methods such as K-Means clustering and supervised methods like the Support Vector Machine (SVM) and Artificial Neural Network (ANN).

Deep Learning Techniques for Improving Digital Gait Segmentation.

Wearable technology for the automatic detection of gait events has recently gained growing interest, enabling advanced analyses that were previously limited to specialist centres and equipment (e.g., instrumented walkway). In this study, we present a novel method based on dilated convolutions for an accurate detection of gait events (initial and final foot contacts) from wearable inertial sensors. A rich dataset has been used to validate the method, featuring 71 people with Parkinson's disease (PD) and 67 healthy control subjects. Multiple sensors have been considered, one located on the fifth lumbar vertebrae and two on the ankles. The aims of this study were: (i) to apply deep learning (DL) techniques on wearable sensor data for gait segmentation and quantification in older adults and in people with PD; (ii) to validate the proposed technique for measuring gait against traditional gold standard laboratory reference and a widely used algorithm based on wavelet transforms (WT); (iii) to assess the performance of DL methods in assessing high-level gait characteristics, with focus on stride, stance and swing related features. The results showed a high reliability of the proposed approach, which achieves temporal errors considerably smaller than WT, in particular for the detection of final contacts, with an inter-quartile range below 70 ms in the worst case. This study showes encouraging results, and paves the road for further research, addressing the effectiveness and the generalization of data-driven learning systems for accurate event detection in challenging conditions.

Design and Evaluation of a Soft Assistive Lower Limb Exoskeleton

SummaryWearable devices are fast evolving to address mobility and autonomy needs of elderly people who would benefit from physical assistance. Recent developments in soft robotics provide important opportunities to develop soft exoskeletons (also called exosuits) to enable both physical assistance and improved usability and acceptance for users. The XoSoft EU project has developed a modular soft lower limb exoskeleton to assist people with low mobility impairments. In this paper, we present the design of a soft modular lower limb exoskeleton to improve person’s mobility, contributing to independence and enhancing quality of life. The novelty of this work is the integration of quasi-passive elements in a soft exoskeleton. The exoskeleton provides mechanical assistance for subjects with low mobility impairments reducing energy requirements between 10% and 20%. Investigation of different control strategies based on gait segmentation and actuation elements is presented. A first hip–knee unilateral prototype is described, developed, and its performance assessed on a post-stroke patient for straight walking. The study presents an analysis of the human–exoskeleton energy patterns by way of the task-based biological power generation. The resultant assistance, in terms of power, was 10.9% ± 2.2% for hip actuation and 9.3% ± 3.5% for knee actuation. The control strategy improved the gait and postural patterns by increasing joint angles and foot clearance at specific phases of the walking cycle.

"Pathological Gait Abnormality Detection and Segmentation by Processing the Hip Joints Motion Data to Support Mobile Gait Rehabilitation"

An accurate detection of the gait sub-phases is fundamental in clinical gait analysis to interpret kinetic and kinematic data. In general, detecting the gait events that mark the transition from one gait sub-phase to another as well as the sequence of sub-phases is essential to evaluate gait abnormalities. However, finding a reliable segmentation for pathological gait has been a challenging task. This manuscript entails a generic approach for the gait segmentation into sub-phases in the CORBYS1 system. A number of distinctive features are extracted from the Hip joints motion data which are able to partition and segment the gait cycles in an efficient way. The degree of deviation (i.e. anomaly) in each sub-phase is then calculated with respect to an optimal gait reference which is used for robot-assisted gait rehabilitation. The proposed gait segmentation method is applicable to gait with many types of pathology since training on the pathology specific templates is not required. Performance of the proposed algorithm is evaluated by statistical analysis of results which produced 100% gait segmentation accuracy for healthy subjects and over 99% for pathological subjects.

Body Sensor Network-Based Gait Quality Assessment for Clinical Decision-Support via Multi-Sensor Fusion

This paper presents a versatile multi-sensor fusion method and decision-making algorithm for ambulatory and continuous patient monitoring purposes via a body sensor network (BSN). Gait features including spatio-temporal parameters, gait asymmetry, and regularity were identified and estimated from individual patients data collected from clinical trials. Hence, a continuous assessment and diagnosis of the improvement or the deterioration of the lower limb rehabilitation process is ensured. The experimental results from 10-m free walking trials indicated that the proposed method has a good consistency with the clinically used observational method. The gait assessment results were comparable with previous studies. Gait segmentation succeed even when the pace deviates significantly from the healthy subjects' reference value, which provides proof of objectivity and effectiveness of this preliminary research, namely, using wearable inertial measurement unit (IMUs) as an indicator to detect gait abnormality in subjects with neurological disorders. The hypothesis of gait quality-related clinical trials were designed and validated via both machine learning approach and feature layer data fusion. With further validations, the proposed inertial sensor-based gait assessment approach has the potential to be applied both routinely in clinical practice and for tele-health scenes such as fall detection of the elder at home.

A Multi-Sensor Matched Filter Approach to Robust Segmentation of Assisted Gait.

Individuals with mobility impairments related to age, injury, or disease, often require the help of an assistive device (AD) such as a cane to ambulate, increase safety, and improve overall stability. Instrumenting these devices has been proposed as a non-invasive way to proactively monitor an individual’s reliance on the AD while also obtaining information about behaviors and changes in gait. A critical first step in the analysis of these data, however, is the accurate processing and segmentation of the sensor data to extract relevant gait information. In this paper, we present a highly accurate multi-sensor-based gait segmentation algorithm that is robust to a variety of walking conditions using an AD. A matched filtering approach based on loading information is used in conjunction with an angular rate reversal and peak detection technique, to identify important gait events. The algorithm is tested over a variety of terrains using a hybrid sensorized cane, capable of measuring loading, mobility, and stability information. The reliability and accuracy of the proposed multi-sensor matched filter (MSMF) algorithm is compared with variations of the commonly employed gyroscope peak detection (GPD) algorithm. Results of an experiment with a group of 30 healthy participants walking over various terrains demonstrated the ability of the proposed segmentation algorithm to reliably and accurately segment gait events.

Segmentation of Gait Sequences in Sensor-Based Movement Analysis: A Comparison of Methods in Parkinson's Disease.

Robust gait segmentation is the basis for mobile gait analysis. A range of methods have been applied and evaluated for gait segmentation of healthy and pathological gait bouts. However, a unified evaluation of gait segmentation methods in Parkinson’s disease (PD) is missing. In this paper, we compare four prevalent gait segmentation methods in order to reveal their strengths and drawbacks in gait processing. We considered peak detection from event-based methods, two variations of dynamic time warping from template matching methods, and hierarchical hidden Markov models (hHMMs) from machine learning methods. To evaluate the methods, we included two supervised and instrumented gait tests that are widely used in the examination of Parkinsonian gait. In the first experiment, a sequence of strides from instructed straight walks was measured from 10 PD patients. In the second experiment, a more heterogeneous assessment paradigm was used from an additional 34 PD patients, including straight walks and turning strides as well as non-stride movements. The goal of the latter experiment was to evaluate the methods in challenging situations including turning strides and non-stride movements. Results showed no significant difference between the methods for the first scenario, in which all methods achieved an almost 100% accuracy in terms of F-score. Hence, we concluded that in the case of a predefined and homogeneous sequence of strides, all methods can be applied equally. However, in the second experiment the difference between methods became evident, with the hHMM obtaining a 96% F-score and significantly outperforming the other methods. The hHMM also proved promising in distinguishing between strides and non-stride movements, which is critical for clinical gait analysis. Our results indicate that both the instrumented test procedure and the required stride segmentation algorithm have to be selected adequately in order to support and complement classical clinical examination by sensor-based movement assessment.

Detecting free-living steps and walking bouts: validating an algorithm for macro gait analysis

Research suggests wearables and not instrumented walkways are better suited to quantify gait outcomes in clinic and free-living environments, providing a more comprehensive overview of walking due to continuous monitoring. Numerous validation studies in controlled settings exist, but few have examined the validity of wearables and associated algorithms for identifying and quantifying step counts and walking bouts in uncontrolled (free-living) environments. Studies which have examined free-living step and bout count validity found limited agreement due to variations in walking speed, changing terrain or task. Here we present a gait segmentation algorithm to define free-living step count and walking bouts from an open-source, high-resolution, accelerometer-based wearable (AX3, Axivity). Ten healthy participants (20–33 years) wore two portable gait measurement systems; a wearable accelerometer on the lower-back and a wearable body-mounted camera (GoPro HERO) on the chest, for 1 h on two separate occasions (24 h apart) during free-living activities. Step count and walking bouts were derived for both measurement systems and compared. For all participants during a total of almost 20 h of uncontrolled and unscripted free-living activity data, excellent relative (rho ⩾ 0.941) and absolute (ICC(2,1) ⩾ 0.975) agreement with no presence of bias were identified for step count compared to the camera (gold standard reference). Walking bout identification showed excellent relative (rho ⩾ 0.909) and absolute agreement (ICC(2,1) ⩾ 0.941) but demonstrated significant bias. The algorithm employed for identifying and quantifying steps and bouts from a single wearable accelerometer worn on the lower-back has been demonstrated to be valid and could be used for pragmatic gait analysis in prolonged uncontrolled free-living environments.

Towards holistic free-living assessment in Parkinson's disease: Unification of gait and fall algorithms with a single accelerometer.

Technological developments have seen the miniaturization of sensors, small enough to be embedded in wearable devices facilitating unobtrusive and longitudinal monitoring in free-living environments. Concurrently, the advances in algorithms have been ad-hoc and fragmented. To advance the mainstream use of wearable technology and improved functionality of algorithms all methodologies must be unified and robustly tested within controlled and free-living conditions. Here we present and unify a (i) gait segmentation and analysis algorithm and (ii) a fall detection algorithm. We tested the unified algorithms on a cohort of young healthy adults within a laboratory. We then deployed the algorithms on longitudinal (7 day) accelerometer-based data from an older adult with Parkinson's disease (PD) to quantify real world gait and falls. We compared instrumented falls to a self-reported falls diary to test algorithm efficiency and discuss the use of unified algorithms to impact free-living assessment in PD where accurate recognition of gait may reduce the number of automated detected falls (38/week). This informs ongoing work to use gait and related outcomes as pragmatic clinical markers.

Hidden Markov model-based strategy for gait segmentation using inertial sensors: Application to elderly, hemiparetic patients and Huntington's disease patients.

A solution to discriminate stance and swing in both healthy and abnormal gait using inertial sensors is proposed. The method is based on a two states hidden Markov model trained in a supervised way. The proposed method can generalize across different groups of subjects, without the need of parameters tuning. Leave-one-subject-out validation tests showed 20 ms and 16 ms errors on average in the determination of foot strike and toe off events across the three groups of subjects including 10 elderly, 10 hemiparetic patients and 10 Huntington's disease patients. The proposed methodology can be implemented online in portable devices to be used in clinical practice or in everyday personal health assessment.

A robust real-time gait event detection using wireless gyroscope and its application on normal and altered gaits

Gait events detection allows clinicians and biomechanics researchers to determine timing of gait events, to estimate duration of stance phase and swing phase and to segment gait data. It also aids biomedical engineers to improve the design of orthoses and FES (functional electrical stimulation) systems. In recent years, researchers have resorted to using gyroscopes to determine heel-strike (HS) and toe-off (TO) events in gait cycles. However, these methods are subjected to significant delays when implemented in real-time gait monitoring devices, orthoses, and FES systems. Therefore, the work presented in this paper proposes a method that addresses these delays, to ensure real-time gait event detection. The proposed algorithm combines the use of heuristics and zero-crossing method to identify HS and TO. Experiments involving: (1) normal walking; (2) walking with knee brace; and (3) walking with ankle brace for overground walking and treadmill walking were designed to verify and validate the identified HS and TO. The performance of the proposed method was compared against the established gait detection algorithms. It was observed that the proposed method produced detection rate that was comparable to earlier reported methods and recorded reduced time delays, at an average of 100 ms.

Providing time-discrete gait information by wearable feedback apparatus for lower-limb amputees: usability and functional validation.

Here we describe a novel wearable feedback apparatus for lower-limb amputees. The system is based on three modules: a pressure-sensitive insole for the measurement of the plantar pressure distribution under the prosthetic foot during gait, a computing unit for data processing and gait segmentation, and a set of vibrating elements placed on the thigh skin. The feedback strategy relies on the detection of specific gait-phase transitions of the amputated leg. Vibrating elements are activated in a time-discrete manner, simultaneously with the occurrence of the detected gait-phase transitions. Usability and effectiveness of the apparatus were successfully assessed through an experimental validation involving ten healthy volunteers.

Estimation of spatio-temporal parameters for post-stroke hemiparetic gait using inertial sensors

This paper represents the first step in developing an inertial sensor system that is capable of assessing post-stroke gait in terms of walking speed and temporal gait symmetry. Two inertial sensors were attached at the midpoint of each shank to measure the accelerations and angular velocity during walking. Despite the abnormalities in hemiparetic gait, the angular velocity of most of the testing subjects (12 out of 13) exhibited similar characteristics as those from a healthy population, enabling walking speed estimation and gait event detection based on the pendulum walking model. The results from a standardized 10-meter walk test demonstrated that the IMU-based method has an excellent agreement with the clinically used stopwatch method. The gait symmetry results were comparable with previous studies. The gait segmentation failed when the angular velocity deviates significantly from the healthy groups’ profile. With further development and concurrent validations, the inertial sensor-based system may eventually become a useful tool for continually monitoring spatio-temporal gait parameters post stroke in a natural environment.

Combining wearable sensors for location-free monitoring of gait in older people

Falls are a major cause of injury and fatality among the older population and the use of wearable sensors to quantify key metrics relating to gait and motor function in order to evaluate potential falls risk are increasing in use. However, one of the problems with current use of quantitative gait data to evaluate falls risk is that it is still tied to a relatively controlled deployment model due to lack of effective methods to label and segment gait data that could be acquired over a long term basis in the uncontrolled home and community setting. This means that we cannot evaluate the potentially powerful impact of environmental factors on gait and motor function in the home and community. In this paper we present a conceptual approach to solving this problem by combining inertial sensing methods for quantitative gait analysis with life logging, using other wearable sensors and a wearable camera that automatically record a wearer's contexts, at all times. Using the system, a clinician can use both gait data and lifelog data to mutually index each other to enable a more thorough exploration of data and a greater understanding of the impact of environment on gait function and subsequent falls risk. We also present a single case study of a long term deployment using a prototype of the system, with feedback from an experienced clinician, to illustrate its potential clinical utility.

Can Triaxial Accelerometry Accurately Recognize Inclined Walking Terrains?

The standard method for the analysis of body accelerations cannot accurately estimate the energy expenditure (EE) of uphill or downhill walking. The ability to recognize the grade of the walking surface will most likely improve upon the accuracy of the EE estimates for daily physical activities. This paper investigates the benefits of automatic gait analysis approaches including step-by-step gait segmentation and heel-strike recognition of the accelerometry signal in classifying various gradients. Triaxial accelerometry signals were collected from 12 subjects, performing walking on seven different gradient surfaces: 1) 92 m of 0(°) flat ground; 2) 85 m of ±2.70(°) inclined ramp; 3) 24 m of ±9.86(°) inclined ramp; and 4) 6-m pitch line of ±28.03(°) rake of stairway. Validity studies performed on a group of randomly selected healthy subjects showed high agreement scores between the automated heel-strike recognition markers, manual gait annotation markers, and video-based gait-segmentation markers. Thirteen subset features were found using a subset-selection search procedure from 57 extracted features which maximize the classification accuracy, performed with a Gaussian mixture model classifier, as estimated using sixfold cross-validation. An overall walking pattern-recognition accuracy of 82.46% was achieved on seven different inclined terrains using the 13 selected features. This system should, therefore, improve the accuracy of daily EE estimates with accurate measures on terrain inclinations.