Distributional semantics and unsupervised clustering for sensor relevancy prediction

Effect of Mirror Therapy on Recovery of Stroke Survivors: A Systematic Review and Network Meta-analysis

BackgroundKinematic analysis has been used to objectively evaluate movement patterns, quality, and strategies during reaching tasks. However, no study has investigated whether kinematic variables during unilateral and bilateral reaching tasks predict a patient’s perceived arm use during activities of daily living (ADL) after an intensive intervention. Therefore, this study investigated whether kinematic measures during unilateral and bilateral reaching tasks before an intervention can predict clinically meaningful improvement in perceived arm use during ADL after intensive poststroke rehabilitation.MethodsThe study was a secondary analysis of 120 subjects with chronic stroke who received 90–120 min of intensive intervention every weekday for 3–4 weeks. Reaching kinematics during unilateral and bilateral tasks and the Motor Activity Log (MAL) were evaluated before and after the intervention.ResultsKinematic variables explained 22 and 11 % of the variance in actual amount of use (AOU) and quality of movement (QOM), respectively, of MAL improvement during unilateral reaching tasks. Kinematic variables also explained 21 and 31 % of the variance in MAL-AOU and MAL-QOM, respectively, during bilateral reaching tasks. Selected kinematic variables, including endpoint variables, trunk involvement, and joint recruitment and interjoint coordination, were significant predictors for improvement in perceived arm use during ADL (P < 0.05).ConclusionsArm–trunk kinematics may be used to predict clinically meaningful improvement in perceived arm use during ADL after intensive rehabilitation. Involvement of interjoint coordination and trunk control variables as predictors in bilateral reaching models indicates that a high level of motor control (i.e., multijoint coordination) and trunk stability may be important in obtaining treatment gains in arm use, especially for bilateral daily activities, in intensive rehabilitation after stroke.

FLAAP: An Open Human Activity Recognition (HAR) Dataset for Learning and Finding the Associated Activity Patterns

Human Activity Recognition Using Ambient Sensor Data

This study aimed to investigate the impacts of combining hand function training with robot-assisted upper-limb rehabilitation on upper-limb function, developmental outcomes, and activities of daily living (ADLs) in children with brain lesions. This single-subject, multiple-baseline study involved 3 children with brain lesions. During the baseline phase, the participants received conventional occupational therapy. In the intervention phase, hand-assisted activities were combined with robot-assisted upper-limb rehabilitation. Each session was evaluated using the InMotion®ARM system. Upper-limb function, developmental status, and ADLs were evaluated before and after the intervention. Visual analysis was performed, and the difference between preintervention and postintervention assessments was calculated to compare outcomes. All participants demonstrated improvements following the intervention, with average increases of 8.33 points in the Fugl-Meyer Assessment for Upper Extremity and 17.25 points in the Pediatric Motor Activity Log. Consistent improvements in personal-social development were also observed based on the Korean Developmental Screening Test for Infants and Children, Second Edition. Session-based assessments using InMotion®ARM indicated enhanced aiming accuracy and reduced movement duration across all participants. Combining hand function with robot-assisted upper-limb rehabilitation positively affected the upper-limb function, developmental progress, and ADLs in children with brain lesions. Future large-scale studies are warranted to confirm these findings and explore their broader clinical applicability.

Combination of Exoskeletal Upper Limb Robot and Occupational Therapy Improve Activities of Daily Living Function in Acute Stroke Patients

Study finds insufficient evidence regarding community provision of hearing aids

The introduction of vision-language models like CLIP has enabled the development of foundational video models capable of generalizing to unseen videos and human actions. However, these models are typically trained on web videos, which often fail to capture the challenges present in Activities of Daily Living (ADL) videos. Existing works address ADL-specific challenges, such as similar appearances, subtle motion patterns, and multiple viewpoints, by combining 3D skeletons and RGB videos. However, these approaches are not integrated with language, limiting their ability to generalize to unseen action classes. In this paper, we introduce SKI models, which integrate 3D skeletons into the vision-language embedding space. SKI models leverage a skeleton-language model, SkeletonCLIP, to infuse skeleton information into Vision Language Models (VLMs) and Large Vision Language Models (LVLMs) through collaborative training. Notably, SKI models do not require skeleton data during inference, enhancing their robustness for real-world applications. The effectiveness of SKI models is validated on three popular ADL datasets for zero-shot action recognition and video caption generation tasks.

A combination of rehabilitation techniques may be useful to optimize individualized treatment and efficient resource utilization. To assess the effect of cognitive therapeutic exercise (CTE) and mirror therapy (MT) combined with task-oriented training (TOT) on upper limb functionality, fine and gross motor skills, use of the paretic arm, functional independence in activities of daily living (ADLs), and quality of life among people with subacute stroke. Randomized trial, longitudinal, and multicenter study with one control group (CG) and two experimental groups (EGs) who were assessed before, after, and 3 mo posttreatment. University of Burgos and Burgos University Hospital, Burgos, Spain. One hundred twenty patients with subacute stroke and motor symptoms in the upper limb recruited through consecutive nonprobabilistic sampling and excluding those with tremor due to another neurological disease or cognitive impairment. MT + TOT (EG1) and CTE + TOT (EG2) over 20 1-hr sessions. T3he CG received usual therapy (TOT) that simulated everyday situations. The studied variables were evaluated with the Action Research Arm Test, Box and Block Test, Motor Activity Log, FIM™-Functional Assessment Measure, and Stroke-Specific Quality of Life Scale. Statistically significant differences were found between the CG group and both EGs, with improvements in upper limb functionality, fine and gross motor skills, use of the paretic arm, functional independence in ADLs, and quality of life. No differences were found between EGs. Application of these combined interventions improved all variables studied. Plain-Language Summary: This study found that the use of cognitive therapeutic exercise and mirror therapy combined with task-oriented training may improve upper limb functionality, fine and gross motor skills, the use of the paretic arm, functional independence in activities of daily living, and quality of life among people with subacute stroke.

E-health systems rely on information and communication technology to support and improve various aspects of health services, delivery, and management. The success of artificial intelligence techniques has led to the emergence of a variety of systems designed to address a wide range of healthcare issues. In particular, gathering data on patient activity and behavior has enabled the development of reliable predictive systems for detecting chronic diseases and forecasting their progression. Human activity detection is a vast and emerging field, and various datasets have been collected for training different machine learning and deep learning (DL) models. The University of Milano Bicocca smartphone-based human activity recognition (UniMiB-SHAR) dataset is widely used for analyzing and recognizing human actions, including walking, running, and other daily activities. However, the autoencoder (AE) technique trained on this dataset yields poor performance. This paper aims to enhance the performance of AEs on the challenging UniMiB-SHAR dataset by introducing a convolutional AE model and employing novel preprocessing techniques, including normalization, magnitude, principal component analysis (PCA), and balancing methods such as SMOTEEN and ADASYNE. The experimental results demonstrate that the proposed AE model achieved successful performance, surpassing the state-of-the-art methods, with accuracies of 96.56% for activities of daily living (ADL), 98.86% for Fall, and 88.47% for the full dataset.

Background: Stroke patients exhibit arm global synkinesis (GS), involuntary movement due to muscle weakness and irregular muscle tension. But currently there are few studies examined the effects of GS on activates of daily living in stroke patients. Objectives: To investigate the effects the effects of task-oriented bilateral movements, which promote brain plasticity and are based on neurological theory, using the unaffected arm and the affected arm. Design: Quasi-randomized trial. Methods: Twenty stroke patients were randomly assigned to experimental group I (n=10) and experimental group II (n=10). Before the intervention, arm GS was measured using surface electromyography, and the Motor Activity Log evaluated the quantitative and qualitative uses of the affected arm in daily life. The same items were measured four weeks later. Results: The changes in the GS of the arm of experimental group I showed statistically significant differences only in bending motions (P<.05). Both groups showed statistically significant differences in the amount of use (AOU) and the quality of movement (QOM) scores (P<.01). Comparing the groups, statistically significant differences in GS appeared during bending motions (P<.05), and in the AOU (P<.01) and the QOM scores (P<.05). Conclusion: The intervention in GS reduced the abnormal muscle tension of the affected side by increasing the use of the ipsilateral motor pathway, indicating its effectiveness in improving upper limb functions with smooth contraction and relaxation of the muscles.

Existing machine learning algorithms for human information inference are typically data-driven models trained on carefully labeled datasets. Given the significant labeling effort, traditional pure data-driven approaches are challenging to implement for emerging smart applications requiring long-term finer-grained information. Taking activities of daily life (ADL) tracking for elders as an example, prior work mostly focused on context-level information learning such as cooking and cleaning. [8]. However, new applications such as evaluating elders' cognitive impairments progress by tracking their ADL engagement requires finer-grained, i.e., action-level information [7]. In practice, labeling the day-length data at such granularity can be very expensive and requires a lot of human efforts [9]. My research focuses on the inference problems in the scope of human physical condition monitoring and activity recognition with limited labeled data. To alleviate the effort of labeling large amounts of data, prior works on semi-supervised learning combine a small amount of labeled data with a large amount of unlabeled data to train the model. However, as the label granularity (number of classes) increasing, the difficulty to distinguish nuance distinctions between finer-grained classes escalates as well. This makes training a robust semi-supervised model for finer-grained classification with less labels difficult if not impossible. Fortunately, coarse-grained (context-level) labels is usually available or cheaper to obtain in practice. In this case, the multi-granularity hierarchy between finer and coarse labels follows the aggregation relation defined in [5]. This hierarchical relation can be leveraged in the tasks of inferring finer-grained information. In addition, it is illustrated by the previous study that co-located multi modality sensing systems capture complementary aspects of the same event [6]. The research question I focus on is how to infer finer-grained human information with coarse-grained labeled data leveraging complementary multi-modal sensing? I target three directions: 1) a cross-granularity semi-supervised setting: how to utilize coarse-grained labeled data with a small amount of finer-grained labeled data to infer finer-grained human information, 2) cross-granularity relationship learning: how to learn the multi-granularity class hierarchy from data and further help the finer-grained human information acquisition, 3) enhancing inference granularity by leveraging multi-model sensing: how to leverage the complimentary co-located multiple sensing modalities to accurately infer finer-grained human information?

Hemiparesis is the main sensorimotor deficit after stroke. It can result in limitations in Activities of Daily Living (ADL) and social participation. Hemiparesis can be treated with behavioral techniques of intensive use of the affected arm, such as constraint-induced movement therapy (CIMT), however, it remains unclear whether motor improvement can lead to increases in the domains of activity and participation. Identify whether CIMT is superior to usual techniques to enhance activity and participation outcomes in stroke survivors. A systematic review with meta-analysis was conducted, based on the PRISMA guidelines. Search databases were: PubMed, LILACS, Embase, SciELO, Cochrane Library, Scopus, Medline, and Web of Science, with no language restriction. Meta-analysis was performed with Review Manager (version 5.3), significance level p ≤ 0.05. A total of 21 articles were included for analysis. Superior effects were observed on motor function and performance in activities of daily living of individuals treated with CIMT. The outcomes measures utilized were: Fugl-Meyer Assessment (p = 0.00001); Wolf motor function test (p = 0.01); Modified Barthel Index (p = 0.00001); Motor Activity log (MAL) Amount of use (AOU) (p = 0.01); MAL Quality of movement (QOM) (p = 0.00001); Action Research Arm Test-ARAT (p = 0.00001); and FIM (p = 0.0007). Our results show that CIMT results in more significant gains in the functional use of the upper limb in ADL and functional independence, demonstrating superior activity and participation results in stroke survivors when compared to conventional therapies.

With the recent spread of mobile devices equipped with different sensors, it is possible to continuously recognise and monitor activities in daily life. This sensor-based human activity recognition is formulated as sequence classification to categorise sequences of sensor values into appropriate activity classes. One crucial problem is how to model features that can precisely represent characteristics of each sequence and lead to accurate recognition. It is laborious and/or difficult to hand-craft such features based on prior knowledge and manual investigation about sensor data. To overcome this, we focus on a feature learning approach that extracts useful features from a large amount of data. In particular, we adopt a simple but effective one, called codebook approach, which groups numerous subsequences collected from sequences into clusters. Each cluster centre is called a codeword and represents a statistically distinctive subsequence. Then, a sequence is encoded as a feature expressing the distribution of codewords. The extensive experiments on different recognition tasks for physical, mental and eye-based activities validate the effectiveness, generality and usability of the codebook approach.

Personality recognition using nonverbal behavioral cues is a challenging task in the Affective Computing field. The majority of existing methods investigate personality assessment in social contexts, such as crowded places or social events, but ignore the role of behaviors as well as personality in nonsocial situations (i.e. during individual activities). In this paper we introduce a novel dataset for behavior understanding and personality recognition in a nonsocial context. Forty-six participants were recorded in an unconstrained indoor space, related to a smart home environment, performing six tasks resembling Activities of Daily Living (ADL). During the experiment, personality scores were collected using self-assessment questionnaires. Furthermore, a temporal framework using a Long-Short Term Memory (LSTM) network is proposed to map nonverbal behavioral features to participants' personality labels. Our experiments showed that nonverbal behaviors are important predictors of personality, confirming theories from the personality psychology field.