Motion Sequences Research Articles

Force-Driven Model for Automated Clear Aligner Staging Design Based on Stepwise Tooth Displacement and Rotation in 3D Space

This study introduced a novel force-driven automated staging design method for clear aligners, aimed at enhancing treatment planning efficiency and outcomes. The method simplified the alignment process into a force-driven mechanics model that calculates forces and moments exerted on teeth while adhering to Newton’s third law, determining their displacement and rotation at each position. An optimal path was generated by iteratively moving teeth from their initial to target positions and subsequently divided into stages based on a predefined step size. The algorithm was implemented in C++ and incorporated into the WebGL-based SmarteeCheck3.0 software for visualization. In a maxillary extraction case, the automated staging method (0.25 mm step size) generated 51 stages in merely 5 s, while manual staging (>0.25 mm step size) necessitated 30 min to achieve 55 stages. In a molar distalization case, the automated method demonstrated similar efficiency advantages, generating 30 stages for the maxilla and 34 for the mandible, compared to 41 stages each in manual staging. The automated staging approach yielded shorter and more precise tooth movement paths that adhered to aligner biomechanics and physical principles, surpassing the limitations of manual staging. For cases requiring entire arch displacement, the method incorporated sequential movements with anchorage control to maintain force equilibrium. This innovative method substantially improved design efficiency and accuracy, ultimately elevating the efficacy of clear aligner therapy, although further biomechanical analyses and experimental validations are needed to refine the model parameters.

Using monitoring and simulation to analyze the failure characteristics of multizone landslides controlled by faults: a case study

Slopes influenced by multiple faults are prone to large-scale landslides triggered by multi-regional failures. Understanding the failure process and sequence is essential for the sustainable development of mining operations. This paper presents a method combining InSAR monitoring and numerical simulation to analyze the failure processes of slopes affected by multiple faults. Using the 298–710 m bench in the central area of the eastern slope of the Nanfen open-pit mine as a case study, the multidimensional small baseline subset (MSBAS) technique of InSAR was employed to calculate the vertical and horizontal deformation time series in the study area, enabling an analysis of the deformation characteristics of multizone bedding sliding. Subsequently, a numerical simulation was performed to replicate the three-dimensional evolution of the multizone failure. The simulation results aligned well with field surveys and InSAR monitoring data. Furthermore, the failure sequence analysis identified critical trigger zones within the region. The findings demonstrate that the proposed method effectively identifies the critical areas and movement sequence of multi-regional failures. In this paper, the faults disrupted the connectivity between slopes, acting as the initial trigger for entire landslide. The first failure occurred in a critical area, creating space for further sliding in other regions and reducing lateral constraints in neighboring areas, eventually leading to significant deformations. Therefore, by using this method to identify potential critical areas before landslides occur, targeted mitigation measures can be implemented to reduce the risk of such events.

Decomposition of a complex motor skill with precise error feedback and intensive training breaks expertise ceiling

Complex motor skills involve intricate sequences of movements that require precise temporal coordination across multiple body parts, posing challenges to mastery based on perceived error or reward. One approach that has been widely used is to decompose such skills into simpler, constituent movement elements during the learning process, thereby aligning the task complexity with the learners’ capacity for accurate execution. Despite common belief and prevalent adoption, the effectiveness of this method remains elusive. Here we addressed this issue by decomposing a sequence of precisely timed coordination of movements across multiple fingers into individual constituent elements separately during piano practice. The results demonstrated that the decomposition training enhanced the accuracy of the original motor skill, a benefit not achieved through mere repetition of movements alone, specifically when skilled pianists received explicit visual feedback on timing error in the order of milliseconds during training. During the training, the patterns of multi-finger movements changed significantly, suggesting exploration of movements to refine the skill. By contrast, neither unskilled pianists who underwent the same training nor skilled pianists who performed the decomposition training without receiving visual feedback on the error showed improved skill through training. These findings offer novel evidences suggesting that decomposing a complex motor skill, coupled with receiving feedback on subtle movement error during training, further enhances motor expertise of skilled individuals by facilitating exploratory refinement of movements.

Lateral End-Range Movement Profile and Shot Effectiveness During Grand Slam Tennis Match-Play.

End-range movements are among the most demanding but least understood in the sport of tennis. Using male Hawk-Eye data from match-play during the 2021-2023 Australian Open tournaments, we evaluated the speed, deceleration, acceleration, and shot quality characteristics of these types of movement in men's Grand Slam tennis. Lateral end-range movements that incorporated a change of direction (CoD) were identified for analysis using k-means (end-range) and random forest (CoD) machine learning models. Peak speed, average deceleration into the CoD, average reacceleration out of the CoD, and the quality of the shot played were computed. Players were grouped based on their ATP rankings (top 10, top 50, and outside top 50) to examine the influence of ranking on movement profiles and shot effectiveness. Our data showed that end-range movements profiles of top 10 and top 50 players were characterized by higher peak speed (d=0.3-0.88), deceleration intensity (d=0.25-0.63), and acceleration intensity (d=0.06-0.51) when compared to players outside the top 50 (p<0.05). Top 10 players also demonstrated greater peak speeds (d=0.59) and acceleration intensities (d=0.45) compared to top 50 players (p<0.05). There was a nonlinear inverse relationship between peak speed and shot quality, such that, as peak speed increased, shot quality decreased-notwithstanding that top 10 players were more likely to hit high-quality shots at higher peak speeds. These results quantify the discrete kinematic characteristics of the sport's most challenging movement sequence and reveal, for the first time, that higher ranked players may possess superior movement potential on court.

Diadochokinetic rate in Thai adults using between auditory-perceptual and instrumental methods

Background: Diadochokinetic (DDK) rate refers to the time taken to execute rapid repetitive movements of articulators during speech. DDK rate assessment has been used to screen or as part of an assessment to diagnose and plan treatment for motor speech disorders. It contained alternating motion rate (AMR) and sequential motion rate (SMR) assessments. There were two main methods for DDK assessment: perceptual speech assessment and instrumental speech assessment. Objective: This study aimed to compare the DDK rate measured by auditoryperceptual and instrumental methods and investigate the effects of age and gender on the DDK rate in normal Thai adults. Materials and methods: In this cross-sectional study, the participants were Thai adults aged 21-78. Sixty participants were divided equally into three age groups: young adult, middle-aged adult, and older adult group. All participants were asked to repeat four speech tasks as rapidly as possible, namely, AMR as [ph a:], [th a:], [kh a:], and SMR as [ph a:-th a:-kh a:]. DDK rates were measured using two methods, namely, the auditory-perceptual method and the instrumental method, which used the Praat program. The paired sample t-test examined the difference in average DDK rates between the two methods. The one-way ANOVA was utilized to analyze the average rate differences among three age groups, while gender differences were analyzed through an independent sample t-test. Results: The averages of SMR between the auditory-perceptual method and thePraat program were significantly different (p&lt;0.05), whereas the averages of AMRwere not significantly different. The findings also found that averages of SMR hadsignificant differences (p&lt;0.05) between three age groups and two genders. The young adult group had a higher average of SMR than middle-aged adult and olderadult groups. Males had a higher average of SMR than females. Conclusion: This result indicates that different assessment methods result in different DDK rates, especially SMR. Therefore, choosing the DDK rate assessment method should be careful and consider its accuracy. Moreover, the SMR task may be more sensitive than the AMR task to age-related change and gender factors.

A Review on Inverse Kinematics, Control and Planning for Robotic Manipulators With and Without Obstacles via Deep Neural Networks

Robotic manipulators are highly valuable tools that have become widespread in the industry, as they can achieve great precision and velocity in pick and place as well as processing tasks. However, to unlock their complete potential, some problems such as inverse kinematics (IK) need to be solved: given a Cartesian target, a method is needed to find the right configuration for the robot to reach that point. Another issue that needs to be addressed when dealing with robotic manipulators is the obstacle avoidance problem. Workspaces are usually cluttered and the manipulator should be able to avoid colliding with objects that could damage it, as well as with itself. Two alternatives exist to do this: a controller can be designed that computes the best action for each moment given the manipulator’s state, or a sequence of movements can be planned to be executed by the robot. Classical approaches to all these problems, such as numeric or analytical methods, can produce precise results but take a high computation time and do not always converge. Learning-based methods have gained considerable attention in tackling the IK problem, as well as motion planning and control. These methods can reduce the computational cost and provide results for every situation avoiding singularities. This article presents a literature review of the advances made in the past five years in the use of Deep Neural Networks (DNN) for IK with regard to control and planning with and without obstacles for rigid robotic manipulators. The literature has been organized in several categories depending on the type of DNN used to solve the problem. The main contributions of each reference are reviewed and the best results are presented in summary tables.

Machine learning models for sequential motion recognition in human activity for logistics operations efficiency improvement

Abstract Human activity recognition (HAR) is a vital component of human–robot collaboration. Recognizing the operational elements involved in an operator’s task is essential for realizing this vision, and HAR plays a key role in achieving this. However, recognizing human activity in an industrial setting differs from recognizing daily living activities. An operator’s activity must be divided into fine elements to ensure efficient task completion. Despite this, there is relatively little related research in the literature. This study aims to develop machine learning models to classify the sequential movement elements of a task. To illustrate this, three logistic operations in an integrated circuit (IC) design house were studied, with participants wearing 13 inertial measurement units manufactured by XSENS to mimic the tasks. The kinematics data were collected to develop the machine learning models. The time series data preprocessing involved applying two normalization methods and three different window lengths. Eleven features were extracted from the processed data to train the classification models. Model validation was carried out using the subject-independent method, with data from three participants excluded from the training dataset. The results indicate that the developed model can efficiently classify operational elements when the operator performs the activity accurately. However, incorrect classifications occurred when the operator missed an operation or awkwardly performed the task. RGB video clips helped identify these misclassifications, which can be used by supervisors for training purposes or by industrial engineers for work improvement.

Combating Voice Spoofing Attacks on Wearables via Speech Movement Sequences

Combating Voice Spoofing Attacks on Wearables via Speech Movement Sequences

마비말장애 환자를 위한 청지각적 평가 시스템(DAS) 개발

Background and Objectives: Dysarthria is a significant motor speech disorder that arises from damage to the central or peripheral nervous system. This damage can result in paralysis and weakness of the muscles responsible for articulation, which negatively affects the speech production process. The objective of this study is to develop a Dysarthria Assessment System (DAS) for the auditory-perceptual evaluation of patients with dysarthria. Method: We conducted an analysis of assessment areas and tasks based on literature published after 2000. This analysis guided the development of assessment content, which was subsequently organized for integration into a web platform. Results: The DAS consists of three main components: Speech Mechanism, Speech, and Speech Motor. The Speech Mechanism component includes questions aimed at assessing the symmetry, range of motion, and strength of the articulatory organs involved in speech production, with items presented separately for rest and activity. The Speech component evaluates respiration, phonation, resonance, articulation, and prosody through tasks such as maximum phonation time, paragraph reading, and picture description. In the Speech Motor component, the system measures the frequency of alternating motion rates and sequential motion rates. Conclusions: The DAS developed in this study is expected to be highly beneficial in clinical practice due to its ability to visualize and automatically calculate results, enabling assessments to be conducted via PC or tablet. Future research will focus on standardizing the DAS through large-scale data collection.

AniClipart: Clipart Animation with Text-to-Video Priors

AbstractClipart, a pre-made graphic art form, offers a convenient and efficient way of illustrating visual content. Traditional workflows to convert static clipart images into motion sequences are laborious and time-consuming, involving numerous intricate steps like rigging, key animation and in-betweening. Recent advancements in text-to-video generation hold great potential in resolving this problem. Nevertheless, direct application of text-to-video generation models often struggles to retain the visual identity of clipart images or generate cartoon-style motions, resulting in unsatisfactory animation outcomes. In this paper, we introduce AniClipart, a system that transforms static clipart images into high-quality motion sequences guided by text-to-video priors. To generate cartoon-style and smooth motion, we first define Bézier curves over keypoints of the clipart image as a form of motion regularization. We then align the motion trajectories of the keypoints with the provided text prompt by optimizing the Video Score Distillation Sampling (VSDS) loss, which encodes adequate knowledge of natural motion within a pretrained text-to-video diffusion model. With a differentiable As-Rigid-As-Possible shape deformation algorithm, our method can be end-to-end optimized while maintaining deformation rigidity. Experimental results show that the proposed AniClipart consistently outperforms existing image-to-video generation models, in terms of text-video alignment, visual identity preservation, and motion consistency. Furthermore, we showcase the versatility of AniClipart by adapting it to generate a broader array of animation formats, such as layered animation, which allows topological changes.

Sub-second characterization of locomotor activities of mouse models of Parkinsonism.

The degeneration of midbrain dopamine (DA) neurons disrupts the neural control of natural behavior, such as walking, posture, and gait in Parkinson's disease. While some aspects of motor symptoms can be managed by dopamine replacement therapies, others respond poorly. Recent advancements in machine learning-based technologies offer opportunities for unbiased segmentation and quantification of natural behavior in both healthy and diseased states. In the present study, we applied the motion sequencing (MoSeq) platform to study the spontaneous locomotor activities of neurotoxin and genetic mouse models of Parkinsonism as the midbrain DA neurons progressively degenerate. We also evaluated the treatment efficacy of levodopa (L-DOPA) on behavioral modules at fine time scales. We revealed robust changes in the kinematics and usage of the behavioral modules that encode spontaneous locomotor activity. Further analysis demonstrates that fast behavioral modules with higher velocities were more vulnerable to loss of DA and preferentially affected at early stages of Parkinsonism. Last, L-DOPA effectively improved the velocity, but not the usage and transition probability, of behavioral modules of Parkinsonian animals. In conclusion, the hypokinetic phenotypes in Parkinsonism are mediated by the decreased velocities of behavioral modules and the disrupted temporal organization of sub-second modules into actions. Moreover, we showed that the therapeutic effect of L-DOPA is mainly mediated by its effect on the velocities of behavior modules at fine time scales. This work documents robust changes in the velocity, usage, and temporal organization of behavioral modules and their responsiveness to dopaminergic treatment under the Parkinsonian state.

Enhancing skeleton-based action recognition with hybrid real and gan-generated datasets

This research addresses the critical challenge of recognizing mutual actions involving multiple individuals, an important task for applications such as video surveillance, human-computer interaction, autonomous systems, and behavioral analysis. Identifying these actions from 3D skeleton motion sequences poses significant challenges due to the necessity of accurately capturing intricate spatial and temporal patterns in diverse, dynamic, and often unpredictable environments. To tackle this, a robust neural network framework was developed that combines Convolutional Neural Networks (CNNs) for efficient spatial feature extraction with Long Short-Term Memory (LSTM) networks to model temporal dependencies over extended sequences. A distinguishing feature of this study is the creation of a hybrid dataset that which combines real-world skeleton motion data with synthetically generated samples, produced using Generative Adversarial Networks (GANs). This dataset enriches variability, enhances generalization, and mitigates data scarcity challenges. Experimental findings across three different network architectures demonstrate that our method significantly enhances recognition accuracy, mainly due to the integration of CNNs and LSTMs alongside the broadened dataset. Our approach successfully identifies complex interactions and ensures consistent performance across different perspectives and environmental conditions. The improved reliability in recognition indicates that this framework can be effectively utilized in practical applications such as security systems, crowd monitoring, and other areas where precise detection of mutual actions is critical, particularly in real-time and dynamic environments

Impact of Neurodynamic Sequencing on the Mechanical Behaviour of the Median Nerve and Brachial Plexus: An Ultrasound Shear Wave Elastography Study.

When performing the Upper Limb Neurodynamic Test 1 (ULNT1), the order of joint movement can be varied to place more stress onto certain nerve segments. However, the mechanisms underlying this phenomenon are still unclear. This study aimed to analyze the differences in the stiffness of the median nerve (MN) and the brachial plexus (BP) using ultrasound shear wave elastography during three sequences of the ULNT1: standard (ULNT1-STD), distal-to-proximal (ULNT1-DIST), and proximal-to-distal (ULNT1-PROX). Shear wave velocity (SWV) was measured at the initial and final position of each sequence at the MN (wrist) and at the C5 and C6 nerve roots (interscalene level) in 31 healthy subjects. A significant interaction was found between ULNT1 sequence and location (p < 0.001). The ULNT1-STD and ULNT1-DIST induced a greater stiffness increase in the MN (5.67 ± 0.91 m/s, +113.94%; 5.65 ± 0.98 m/s, +115.95%) compared to C5 and C6 (p < 0.001). The ULNT1-PROX resulted in a significantly smaller increase in stiffness at the MN (4.13 ± 0.86 m/s, +54.17%, p < 0.001), but a greater increase at C5 (4.88 ± 1.23 m/s, +53.39%, p < 0.001) and at C6 (4.87 ± 0.81 m/s, +31.55%). The differences for the ULNT1-PROX at C6 were only significant compared to the ULNT1-STD (p < 0.001), but not the ULNT1-DIST (p = 0.066). BP and MN stiffness vary depending on the joint movement sequence during neurodynamic testing. However, the influence of the surrounding tissues may have affected SWV measurements; consequently, these results should be interpreted with caution.

Effect of Pilates and Yoga on Balance in Subjects with Hemiparetic Stroke – A Comparative Pilot Study

Background: In patients with stroke, balance is mainly affected in day-to-day life activity. Pilates and yoga include physical and mental components focusing on breathing control, spinal alignment and flexibility via multiple exercises. Pilates focuses on core stability and spinal alignment, whereas yoga combines specific postures, movement sequences, relaxation and meditation. Aim: To compare the effect of Pilates and Yoga on balance in patient with hemiparetic stroke. Material &amp; Method: This study was conducted in various neurological set ups of Surat. 10 hemiparetic stroke patients were randomly assigned into 2 groups: 5 in group A (Pilates &amp; conventional physiotherapy exercise) and 5 in group B (Yoga &amp; conventional physiotherapy exercise). Patients in both groups received intervention for 3 days/week for 45 minutes for 4 weeks. Conventional physiotherapy was given 6 days/week. Balance was assessed by using Berg Balance Scale. Data were recorded at baseline and after 4 weeks of training. Descriptive statistics was calculated using mean and standard deviation, Paired t-test was used to find difference within group and independent t-test for between group comparison. Result: The results showed there was no significant difference in between group on balance but there was improvement in balance in both groups. Conclusion: After 4 weeks of training, there was significant improvement in balance in both groups. And there was no significant difference in pilates versus yoga in patients with hemiparetic stoke. KEYWORDS: Balance, Hemiparetic stroke, Pilates, Yoga.

Exploiting travel sequences to optimise facility layouts with multiple input/output points

The facility layout problem (FLP) involves arranging departments on a shop floor to optimise specific objectives, traditionally focussing on pairwise flows between departments. However, these methods often underestimate total travel distances, especially when flows involve multiple input/output points and visits to more than two departments. To address this, connected movements – actual routes taken by transporters – must be considered. This study uses data captured from an Internet of Things (IoT) network and stored on cloud servers to analyze worker movements and accurately calculate travel distances. A mixed-integer programming model is proposed to minimise total travel distance using connected movements as input. Due to the problem's complexity, a biased random key genetic algorithm is employed to find optimal layouts. A case study at a fertiliser production company demonstrates the effectiveness of the approach, achieving a 15% reduction in travel distance compared to layouts generated by traditional methods. The IoT-enabled method also minimises productivity losses by optimising worker movements. While the study focuses on fertiliser manufacturing, the findings are applicable to other settings, such as warehousing, where complex movement sequences and multiple IO points are common in processes like picking, packing, and shipping.

Autoregressive GAN for Semantic Unconditional Head Motion Generation

In this work, we address the task of unconditional head motion generation to animate still human faces in a low-dimensional semantic space from a single reference pose. Different from traditional audio-conditioned talking head generation that seldom puts emphasis on realistic head motions, we devise a GAN-based architecture that learns to synthesize rich head motion sequences over long duration while maintaining low error-accumulation levels. In particular, the autoregressive generation of incremental outputs ensures smooth trajectories, while a multi-scale discriminator on input pairs drives generation toward better handling of high- and low-frequency signals and less mode collapse. We experimentally demonstrate the relevance of the proposed method and show its superiority compared to models that attained state-of-the-art performances on similar tasks.

Using High-Speed Videoendoscopy to Analyze Laryngeal Closure Parameters During Normal Swallow.

This pilot study was designed to test the tolerability of a lower scope position and feasibility of custom-designed MATLAB graphical user interface (GUI) used to analyze playback review of laryngeal high-speed videoendoscopy (laryngeal HSV) during healthy volitional dry swallows. We hypothesized this method would conceptually provide time resolution for glottic closure events compared with standard (30 frames per second, fps), and enable a means to measure timing, sequence, and duration of laryngeal movements during swallowing not otherwise visualized. A total of 14 healthy adults (4 male, 22-80 years) participated. We performed laryngeal HSV at 500fps. Measurements included: (i) feasibility and tolerability of the procedure; (ii) identification of a swallowing segment of interest (SOI) for the peak of the swallow; and (iii) description of laryngeal swallowing movements using a GUI. Fourteen subjects tolerated the procedure without discomfort and swallow images were able to be analyzed in 12. Using our GUI, mean SOI was 260 ms, yielding 130 frames for analysis (compared with seven in standard laryngoscopy). Vocal fold adduction, vocal fold medialization, and anterior-posterior arytenoid compression to the epiglottis prior to whiteout could be identified and sequenced. Participants tolerated a low position of the endoscope during dry volitional swallows. The output of our GUI demonstrated a novel technique for identifying, describing, and sequencing a swallowing SOI. Future studies should investigate laryngeal closure and arytenoid positioning with a bolus and in a range of ages, genders, and etiologies in both healthy and abnormal populations to better understand swallowing physiology. NA Laryngoscope, 2024.

Object-based feedback attention in convolutional neural networks improves tumour detection in digital pathology

Human visual attention allows prior knowledge or expectations to influence visual processing, allocating limited computational resources to only that part of the image that are likely to behaviourally important. Here, we present an image recognition system based on biological vision that guides attention to more informative locations within a larger parent image, using a sequence of saccade-like motions. We demonstrate that at the end of the saccade sequence the system has an improved classification ability compared to the convolutional neural network (CNN) that represents the feedforward part of the model. Feedback activations highlight salient image features supporting the explainability of the classification. Our attention model deviates substantially from more common feedforward attention mechanisms, which linearly reweight part of the input. This model uses several passes of feedforward and backward activation, which interact non-linearly. We apply our feedback architecture to histopathology patch images, demonstrating a 3.5% improvement in accuracy (p < 0.001) when retrospectively processing 59,057 9-class patches from 689 colorectal cancer WSIs. In the saccade implementation, overall agreement between expert-labelled patches and model prediction reached 93.23% for tumour tissue, surpassing inter-pathologist agreement. Our method is adaptable to other areas of science which rely on the analysis of extremely large-scale images.

Mu down regulation EEG-neurofeedback training combined to motor imagery facilitates early consolidation in a sequential finger tapping task

Objective.Motor imagery (MI) has demonstrated positive effects on motor performance and triggers activation in the motor cortex (MC). EEG-Neurofeedback (EEG-NF) is a neuromodulation technique that provides real-time feedback on one's brain activity, enabling self-regulation of brain states. While there is increasing evidence of humans controlling the activity of various brain networks, including the MC, through EEG-NF, the tangible benefits of this self-regulation on motor performance remain uncertain. This study investigates the potential benefits of EEG-NF training in explicit learning of a sequential movement, in comparison to MI training and to a combined EEG-NF and MI training.Approach.Ninety-one right-handed healthy adults were randomly assigned to one of four groups (a)NF(n= 24), (b)MI(n= 22), (c)MI + NF(n= 23) and (d)control(n= 22). Participants performed a sequential finger tapping task before and after (immediately, 20 min and 24 h) a single 30 min training session. Motor performance, movement speed and event related desynchronization data were analyzed.Main results.MI training led to a better motor performance compared to control condition immediately after training that was sustained at the 20 min retest time point (p= 0.02 and 0.05). In contrast, EEG-NF training alone did not yield better motor performance compared to control condition at any time-point (p> .05). Remarkably, only the combination of both trainings led to superior motor performance 24 h after training in comparison to control group (p= 0.02). Additionally, all experimental groups successfully decreased mu rhythm amplitude throughout most of the training.Significance.Combined EEG-NF and MI training appears particularly promising for enhancing motor consolidation holding the potential to advance rehabilitation approaches.

Patient activity recognition for robust performance under varied illumination and cluttered background in indoor environment

Human Activity Recognition (HAR) holds significant importance in health and human-machine interaction. However, recognizing actions from 2D information faces challenges like occlusion, illumination variation, cluttered backgrounds, and view invariance. These hurdles are particularly pronounced in indoor patient monitoring settings due to fluctuating lighting conditions and cluttered backgrounds, which compromise the accuracy of activity recognition systems. A new architecture named illuminationRevive has been proposed to tackle this issue. Which utilizes an encoder-decoder convolutional neural network and image post-processing blocks to enhance the image's visual appearance. A new dataset comprising seven indoor physical activities has been proposed, created with contributions from thirty individuals aged 20 to 45. A hybrid fusion architecture is proposed to classify activities, integrating motion sequence information and body joint features. The proposed classification model incorporates generated Skeleton Motion History Images (SMHIs), collected human joint motion features from video frames, and novel kinematic and geometric features within window frames as inputs. The model can extract spatial and temporal feature vectors by integrating ResNet50-ViT (Residual Network-50 layers, Vision Transformer) and CNN-BiLSTM (Convolutional Neural Network-Bidirectional Long Short-Term Memory) layers. The suggested classification model was evaluated alongside state-of-the-art models using the LNMIIT-SMD (The LNM Institute of Information Technology-Skelton Motion Dataset) and established NTU-RGBD (Nanyang Technological University's Red Blue Green and Depth information) dataset. The evaluation aimed to assess the effectiveness of the proposed classification model architecture. Results demonstrate the superiority of the proposed model, achieving impressive accuracies of 98.21% on real-time data, 98.45% on the proposed dataset, and 97.12% on the NTU-RGBD dataset.