Matteo Giuberti

Automatic UPDRS Evaluation in the Sit-to-Stand Task of Parkinsonians: Kinematic Analysis and Comparative Outlook on the Leg Agility Task

In this study, we first characterize the sit-to-stand (S2S) task, which contributes to the evaluation of the degree of severity of the Parkinsons disease (PD), through kinematic features, which are then linked to the Unified Parkinsons disease rating scale (UPDRS) scores. We propose to use a single body-worn wireless inertial node placed on the chest of a patient. The experimental investigation is carried out considering 24 PD patients, comparing the obtained results directly with the kinematic characterization of the leg agility (LA) task performed by the same set of patients. We show that i) the S2S and LA tasks are rather unrelated and ii) the UPDRS distributions (for both S2S and LA tasks) across the patients have a direct impact on the observed system performance.

Estimation of full-body poses using only five inertial sensors: an eager or lazy learning approach?

Human movement analysis has become easier with the wide availability of motion capture systems. Inertial sensing has made it possible to capture human motion without external infrastructure, therefore allowing measurements in any environment. As high-quality motion capture data is available in large quantities, this creates possibilities to further simplify hardware setups, by use of data-driven methods to decrease the number of body-worn sensors. In this work, we contribute to this field by analyzing the capabilities of using either artificial neural networks (eager learning) or nearest neighbor search (lazy learning) for such a problem. Sparse orientation features, resulting from sensor fusion of only five inertial measurement units with magnetometers, are mapped to full-body poses. Both eager and lazy learning algorithms are shown to be capable of constructing this mapping. The full-body output poses are visually plausible with an average joint position error of approximately 7 cm, and average joint angle error of 7∘. Additionally, the effects of magnetic disturbances typical in orientation tracking on the estimation of full-body poses was also investigated, where nearest neighbor search showed better performance for such disturbances.

Inertial BSN-Based Characterization and Automatic UPDRS Evaluation of the Gait Task of Parkinsonians

The analysis and assessment of motor tasks, such as gait, can provide important information on the progress of neurological disorders such as Parkinsons disease (PD). In this paper, we design a Boby Sensor Network (BSN)-based system for the characterization of gait in Parkinsonians through the extraction of kinematic features, in both time and frequency domains, embedding information on the status of the PD. The gait features extraction is performed on a set of 34 PD patients using a BSN formed by only three inertial nodes (one on the chest and one per thigh). We investigate also the relationship between the selected kinematic features and the Unified Parkinsons Disease Rating Scale (UPDRS) scores assigned to patients by expert neurologists. This work extends a previously proposed approach to the analysis of leg agility and sit-to-stand tasks and, as such, represents a further step to develop a system for automatic and comprehensive evaluation of different PD motor tasks. A performance analysis of different classification techniques is carried out, showing the feasibility of an automatic (and, eventually, remote) UPDRS scoring system, suitable for tele-health applications in the realm of affective medicine.

Fingerprinting-based wireless 3D localization for motion capture applications

In this paper, we consider a radio fingerprinting-based localization system for indoor motion capture applications. Fingerprinting allows target localization on the basis of radio-frequency measurements of the Received radio Signal Strength (RSS), taking into account the presence of fading by means of a training phase. Motion capture is then performed by localizing, through fingerprinting, a group of targets placed on the portion of interest of the user arm---the approach can be easily extended to other portions of the user body. We experimentally investigate, through a SunSPOT wireless sensor network test-bed, different fingerprinting-based localization algorithms, namely deterministic and probabilistic, optimizing, in each case, the system parameters. In particular, the optimization is carried out by minimizing the localization error.

Simple and robust BSN-based activity classification: winning the first BSN contest

Wireless sensor networks (WSNs) are becoming more and more attractive because of their flexibility. In particular, WSNs are being applied to a user body in order to monitor and detect some activities of daily living (ADL) performed by the user (e.g., for medical purposes). This class of WSNs are typically denoted as body sensor networks (BSNs). In this paper, we present a simple, yet accurate and robust, BSN-based activity classification algorithm that can detect and classify a sequence of activities, chosen from a limited set of fixed known activities, by observing the outputs generated by accelerometers and gyroscopes at the sensors placed over the body. This approach has led us to win the first BSN contest [1] and the presented results refer to the experimental data (publicly available) provided in this contest.

Estimation of vertical ground reaction forces and sagittal knee kinematics during running using three inertial sensors

Analysis of running mechanics has traditionally been limited to a gait laboratory using either force plates or an instrumented treadmill in combination with a full-body optical motion capture system. With the introduction of inertial motion capture systems, it becomes possible to measure kinematics in any environment. However, kinetic information could not be provided with such technology. Furthermore, numerous body-worn sensors are required for a full-body motion analysis. The aim of this study is to examine the validity of a method to estimate sagittal knee joint angles and vertical ground reaction forces during running using an ambulatory minimal body-worn sensor setup. Two concatenated artificial neural networks were trained (using data from eight healthy subjects) to estimate the kinematics and kinetics of the runners. The first artificial neural network maps the information (orientation and acceleration) of three inertial sensors (placed at the lower legs and pelvis) to lower-body joint angles. The estimated joint angles in combination with measured vertical accelerations are input to a second artificial neural network that estimates vertical ground reaction forces. To validate our approach, estimated joint angles were compared to both inertial and optical references, while kinetic output was compared to measured vertical ground reaction forces from an instrumented treadmill. Performance was evaluated using two scenarios: training and evaluating on a single subject and training on multiple subjects and evaluating on a different subject. The estimated kinematics and kinetics of most subjects show excellent agreement (ρ>0.99) with the reference, for single subject training. Knee flexion/extension angles are estimated with a mean RMSE <5°. Ground reaction forces are estimated with a mean RMSE < 0.27 BW. Additionaly, peak vertical ground reaction force, loading rate and maximal knee flexion during stance were compared, however, no significant differences were found. With multiple subject training the accuracy of estimating discrete and continuous outcomes decreases, however, good agreement (ρ > 0.9) is still achieved for seven of the eight different evaluated subjects. The performance of multiple subject learning depends on the diversity in the training dataset, as differences in accuracy were found for the different evaluated subjects.

BSN-Based Activity Classification: A Low Complexity Windowing-&-Classification Approach

Wireless sensor networks (WSNs) are becoming more and more attractive because of their flexibility. In particular, WSNs are being applied to a user body in order to monitor and detect some activities of daily living (ADL) performed by the user (e.g., for medical purposes). This class of WSNs are typically denoted as body sensor networks (BSNs). In this paper, we discuss BSN-based human activity classification. In particular, the goal of our approach is to detect a sequence of activities, chosen from a limited set of fixed known activities, by observing the outputs generated by accelerometers and gyroscopes at the sensors placed over the body. In general, our framework is based on low-complexity windowing-&-classification. First, we consider the case of disjoint (in the time domain) activities; then, we extend our approach to encompass a scenario with consecutive non-disjoint activities. While in the first case windowing is separate from classification, in the second case windowing and classification need to be carried out jointly. The obtained results show a significant detection accuracy of the proposed method, making it suitable for healthcare monitoring applications.

A low-complexity activity classification algorithm with optimized selection of accelerometric features

Activity classification consists in detecting and classifying a sequence of activities, choosing from a limited set of known activities, by observing the outputs generated by (typically) inertial sensor devices placed over the body of a user. To this end, machine learning techniques can be effectively used to detect meaningful patterns from data without explicitly defining classification rules. In this paper, we present a novel Body Sensor Network (BSN)-based low complexity activity classification algorithm, which can effectively detect activities performed by the user just analyzing the accelerometric signals generated by the BSN. A preliminary (computationally intensive) training phase, performed once, is run to automatically optimize the key parameters of the algorithm used in the following (computationally light) online phase for activity classification. In particular, during the training phase, optimized subsets of nodes are selected in order to minimize the number of relevant features and keep a good compromise between performance and time complexity. Our results show that the proposed algorithm outperforms other known activity classification algorithms, especially when using a limited number of nodes, and lends itself to real-time implementation.

On the correlation between UPDRS scoring in the leg agility, sit-to-stand, and gait tasks for parkinsonians

Recently, we have proposed a unified approach, based on the use of a Body Sensor Network (BSN) formed by a few body-worn wireless inertial nodes, for automatic assignment of Unified Parkinsons Disease Rating Scale (UPDRS) scores in the following tasks: Leg Agility (LA), Sit-to-Stand (S2S), and Gait (G). Unlike our previous works and the majority of the works appeared in the literature, where UPDRS tasks are investigated singularly, in the current paper we carry out a comparative investigation of the LA, S2S, and G tasks. In particular, we focus on the correlation between UPDRS values assigned to the three tasks by both an expert neurologist and our automatic system. We also consider an aggregate UPDRS score in order to highlight the relevance of each task in the assessment of the gravity of the Parkinson;s Disease (PD).

Motion Capture: From Radio Signals to Inertial Signals

The study of the motion of individuals allows to gather relevant information on a person status, to be used in several fields (e.g., medical, sport, and entertainment). Over the past decade, the research activity in motion capture has benefited from the progress of portable and mobile sensors, paving the way toward the use of motion capture techniques in mHealth applications (e.g., remote monitoring of patients, and telerehabilitation). Indeed, even if the optical motion capture, which typically relies on a set of fixed cameras and body-worn reflecting markers, is generally perceived as the standard reference approach, other motion capture techniques, such as radio and inertial, are attracting an increasing attention because of their suitability in remote mHealth applications.