Dynamic Time Warping Algorithm Research Articles

Enhanced Design of a Tai Chi Teaching Assistance System Integrating DTW Algorithm and SVM

Physical education using technology has enabled traditional practices like Tai Chi, a martial art known for its multiple health benefits and meditative aspects, to set coordinated goals. This research presents an intelligent Tai Chi Teaching Assistance System supported by the integration of the Dynamic Time Warping algorithm and Support Vector Machine, in which can practitioners providing real-time feedback to improve Tai Chi learning and quality. In the system, the DTWA Dynamic Time Warping Algorithm was used to accurately compare a practitioner’s complex body movements with the Tai Chi standard movements dataset, taking into account execution speed deviations and others. Meanwhile, the SVM was employed to classify the movement as to quality and correctness, thereby being able to provide precise, individual feedback. This hybrid approach ensures a high-motion recognition accuracy rate while also adhering to nuanced Tai Chi requirements. The system was evaluated through detailed testing with various levels of Tai Chi experience. Evaluation showed that the students’ performance and understanding of most Taijiquan movements and related physical exercises improved significantly. It indicates the system has a practical application value for also beginners and intermediate and last expert, respectively. It also shows the effectiveness of combining DTW and SVM to support learners ‘body movement trajectory in a physical learning environment, opening them up to additional technology-assisted physical training applications. This provides implications for a more promising generation of future physical education involving the incorporation of complex AI technology.

A novel hybrid database security management technique

Since the advent of the internet and the digitization of information, more individuals and organizations have gained access to the internet, leading to a significant expansion in the size of our database systems. These databases are relied upon for day-to-day activities by individuals and organizations. Furthermore, the advancements in technology have increased the vulnerability of database management systems and enforcing adequate database security measures is of paramount importance to individuals and organizations. In this study, we deployed a hybrid approach to database security system that integrates two techniques: dynamic time-warping algorithm and voice recognition methods. The implementation follows Object-Oriented Analysis and Design procedures, utilizing SQL Management Studio 2014 for the backend and ASP.NET C# for the frontend. Our hybrid approach demonstrates enhanced database security compared to previously employed authentication measures.

RGB 이미지 기반 인간 동작 추정을 통한 투구 동작 분석

Pitching is a major part of baseball, so much so that it can be said to be the beginning of baseball. Analysis of accurate pitching motions is very important in terms of performance improvement and injury prevention. When analyzing the correct pitching motion, the currently used motion capture method has several critical environmental drawbacks. In this paper, we propose analysis of pitching motion using the RGB-based Human Pose Estimation (HPE) model to replace motion capture, which has these shortcomings, and use motion capture data and HPE data to verify its reliability. The similarity of the two data was verified by comparing joint coordinates using the Dynamic Time Warping (DTW) algorithm.

RFM user value tags and XGBoost algorithm for analyzing electricity customer demand data

With the increasing demand for electricity, predicting user electricity demand has become an essential task. The electricity demand characteristics of users in the electricity market are different, so it is necessary to classify and predict users. Aiming at the above problems, a classified forecasting model of electricity demand based on recent consumption, frequency, monetary (RFM), K-means, XGBoost and dynamic time warping (DTW) algorithm is proposed. The experiment showcases that among the electricity consumption of commercial users, the first type of load has the lowest proportion in autumn, at around 18.6 %; The second type of load has the highest proportion in autumn, about 81.3 %; Accurate classification has been made for the consuming quantity of electricity of commercial users. The average error in the forecasting results of the RFM-KM-XGboost model and the actual value of commercial electricity demand is about 0.07 kW; The average errors between the forecasting results of SVM model and RF model and the true values are about 0.2 kW and 0.14 kW, respectively; It indicates that the forecasting error of the RFM-KM-XGBoost model is smaller. The above results indicate that the RFM-KM-XGBoost model can extract users' electricity demand characteristics by classifying user types and load types, and make more accurate predictions of electricity demand for different types of users.

Monitoring Plant Height and Spatial Distribution of Biometrics with a Low-Cost Proximal Platform.

Measuring canopy height is important for phenotyping as it has been identified as the most relevant parameter for the fast determination of plant mass and carbon stock, as well as crop responses and their spatial variability. In this work, we develop a low-cost tool for measuring plant height proximally based on an ultrasound sensor for flexible use in static or on-the-go mode. The tool was lab-tested and field-tested on crop systems of different geometry and spacings: in a static setting on faba bean (Vicia faba L.) and in an on-the-go setting on chia (Salvia hispanica L.), alfalfa (Medicago sativa L.), and wheat (Triticum durum Desf.). Cross-correlation (CC) or a dynamic time-warping algorithm (DTW) was used to analyze and correct shifts between manual and sensor data in chia. Sensor data were able to reproduce with minor shifts in canopy profile and plant status indicators in the field when plant heights varied gradually in narrow-spaced chia (R2 = 0.98), faba bean (R2 = 0.96), and wheat (R2 = up to 0.99). Abrupt height changes resulted in systematic errors in height estimation, and short-scale variations were not well reproduced (e.g., R2 in widely spaced chia was 0.57 to 0.66 after shifting based on CC or DTW, respectively)). In alfalfa, ultrasound data were a better predictor than NDVI (Normalized Difference Vegetation Index) for Leaf Area Index and biomass (R2 from 0.81 to 0.84). Maps of ultrasound-determined height showed that clusters were useful for spatial management. The good performance of the tool both in a static setting and in the on-the-go setting provides flexibility for the determination of plant height and spatial variation of plant responses in different conditions from natural to managed systems.

Electric Energy Metering Error Evaluation Method Based on Deep Learning

The measuring accuracy of the electric energy meter, voltage transformer and current transformer shows a dynamic state under the influence of its factors and external factors. The error of the voltage transformer and current transformer cannot be measured by traditional method. This paper establishes a multidimensional error analysis and fault diagnosis system for power metering based on Hadoop architecture and Spark memory calculation. The platform extracted the error signal from the measurement data and calculated the characteristic value of the error signal. Then, dependent cloud and dynamic time rules are used to estimate the transformer's and voltage transformer's continuity. Then, a half-step membership degree cloud generation algorithm is constructed to overcome the error bias randomness and fuzzy characteristics under the influence factors. Finally, the system uses the dynamic correction method to estimate the similarity of error timing and quantitative factors to realize the error calculation of the current transformer and voltage transformer. The power metering error processing system was built with the support of Hadoop and Spark. The timing increment is introduced in the process of data collection. Dependent cloud and dynamic time-repair methods can improve the accuracy of diagnosing errors in electric energy metering. The parallel optimization of big data platforms by belonging to the cloud and dynamic time-warping algorithm is verified.  

Spatial-temporal graph neural ODE networks for skeleton-based action recognition.

In the field of skeleton-based action recognition, accurately recognizing human actions is crucial for applications such as virtual reality and motion analysis. However, this task faces challenges such intraindividual action differences and long-term temporal dependencies. To address these challenges, we propose an innovative model called spatial-temporal graph neural ordinary differential equations (STG-NODE). First, in the data preprocessing stage, the dynamic time warping (DTW) algorithm is used to normalize and calculate 3D skeleton data to facilitate the derivation of customized adjacency matrices for improving the influence of intraindividual action differences. Secondly, a custom ordinary differential equation (ODE) integrator is applied based on the initial conditions of the temporal features, producing a solution function that simulates the dynamic evolution trend of the events of interest. Finally, the outstanding ODE solver is used to numerically solve the time features based on the solution function to increase the influence of long-term dependencies on the recognition accuracy of the model and provide it with a more powerful temporal modeling ability. Through extensive experiments conducted on the NTU RGB+D 60 and Kinetics Skeleton 400 benchmark datasets, we demonstrate the superior performance of STG-NODE in the action recognition domain. The success of the STG-NODE model also provides new ideas and methods for the future development of the action recognition field.

Research on Fault Detection by Flow Sequence for Industrial Internet of Things in Sewage Treatment Plant Case.

Classifying the flow subsequences of sensor networks is an effective way for fault detection in the Industrial Internet of Things (IIoT). Traditional fault detection algorithms identify exceptions by a single abnormal dataset and do not pay attention to the factors such as electromagnetic interference, network delay, sensor sample delay, and so on. This paper focuses on fault detection by continuous abnormal points. We proposed a fault detection algorithm within the module of sequence state generated by unsupervised learning (SSGBUL) and the module of integrated encoding sequence classification (IESC). Firstly, we built a network module based on unsupervised learning to encode the flow sequence of the different network cards in the IIoT gateway, and then combined the multiple code sequences into one integrated sequence. Next, we classified the integrated sequence by comparing the integrated sequence with the encoding fault type. The results obtained from the three IIoT datasets of a sewage treatment plant show that the accuracy of the SSGBUL-IESC algorithm exceeds 90% with subsequence length 10, which is significantly higher than the accuracies of the dynamic time warping (DTW) algorithm and the time series forest (TSF) algorithm. The proposed algorithm reaches the classification requirements for fault detection for the IIoT.

Sensitivity of Electrocardiogram on Electrode-Pair Locations for Wearable Devices: Computational Analysis of Amplitude and Waveform Distortion.

An electrocardiogram (ECG) is used to observe the electrical activity of the heart via electrodes on the body surface. Recently, an ECG with fewer electrodes, such as a bipolar ECG in which two electrodes are attached to the chest, has been employed as wearable devices. However, the effect of different geometrical factors and electrode-pair locations on the amplitude and waveform of ECG signals remains unclear. In this study, we computationally evaluated the effects of body morphology, heart size and orientation, and electrode misalignment on ECG signals for 48 scenarios using 35 bipolar electrode pairs (1680 waveforms) with a dynamic time warping (DTW) algorithm. It was observed that the physique of the human body model predominantly affected the amplitude and waveform of the ECG signals. A multivariate analysis indicated that the heart-electrode distance and the solid angle of the heart from the electrode characterized the amplitude and waveform of the ECG signals, respectively. Furthermore, the electrode locations for less individual variability and less waveform distortion were close to the location of electrodes V2 and V3 in the standard 12-lead. These findings will facilitate the placement of ECG electrodes and interpretation of the measured ECG signals for wearable devices.

WPS:A whole phenology-based spectral feature selection method for mapping winter crop from time-series images

Accurately obtaining the spatial distribution and planting patterns of crops is very important for agricultural planning and food security. At present, time-series images have been proved to be an effective tool to characterize crop seasonal growth patterns, and identifying crop information by measuring the time-series similarity between unknown classes and known crop phenology curves is also considered to be a useful way. However, the existing methods of selecting feature ignore the connection between each phenological stage of crops and the unique growth rules of the whole phenology, which makes it difficult to select time-series spectral features that are potentially important for crop mapping. In order to make up for this problem, a Whole Phenology-based Spectral Feature Selection (WPS) method was proposed. The aim was to select the time-series feature sets with great differences among winter crops from a large number of spectral features, so as to improve the mapping accuracy of winter rapeseed and winter wheat. Firstly, spectral separability between all classes is calculated. Secondly, the key phenological periods of winter crops were selected according to the importance of temporal features, and the spectral feature sets with high separability were selected according to the key phenological periods. Finally, a Time-weighted Dynamic Time Warping (TWDTW) algorithm was used to generate the winter rapeseed and winter wheat maps of two cities in the middle and lower reaches of the Yangtze River. The mapping accuracy of the two crops is more than 92%, which matches the crop planting area well. The research shows that combining the WPS method with the TWDTW mapping method has great potential to accurately map crop types based on satellite time-series images.

Development of Wearable Devices for Collecting Digital Rehabilitation/Fitness Data from Lower Limbs.

Lower extremity exercises are considered a standard and necessary treatment for rehabilitation and a well-rounded fitness routine, which builds strength, flexibility, and balance. The efficacy of rehabilitation programs hinges on meticulous monitoring of both adherence to home exercise routines and the quality of performance. However, in a home environment, patients often tend to inaccurately report the number of exercises performed and overlook the correctness of their rehabilitation motions, lacking quantifiable and systematic standards, thus impeding the recovery process. To address these challenges, there is a crucial need for a lightweight, unbiased, cost-effective, and objective wearable motion capture (Mocap) system designed for monitoring and evaluating home-based rehabilitation/fitness programs. This paper focuses on the development of such a system to gather exercise data into usable metrics. Five radio frequency (RF) inertial measurement unit (IMU) devices (RF-IMUs) were developed and strategically placed on calves, thighs, and abdomens. A two-layer long short-term memory (LSTM) model was used for fitness activity recognition (FAR) with an average accuracy of 97.4%. An intelligent smartphone algorithm was developed to track motion, recognize activity, and calculate key exercise variables in real time for squat, high knees, and lunge exercises. Additionally, a 3D avatar on the smartphone App allows users to observe and track their progress in real time or by replaying their exercise motions. A dynamic time warping (DTW) algorithm was also integrated into the system for scoring the similarity in two motions. The system's adaptability shows promise for applications in medical rehabilitation and sports.

Enhancing Image Alignment in Time-Lapse-Ground-Penetrating Radar through Dynamic Time Warping

Ground-penetrating radar (GPR) is a rapid and non-destructive geophysical technique widely employed to detect and quantify subsurface structures and characteristics. Its capability for time lapse (TL) detection provides essential insights into subsurface hydrological dynamics, including lateral flow and soil water distribution. However, during TL-GPR surveys, field conditions often create discrepancies in surface geometry, which introduces mismatches across sequential TL-GPR images. These discrepancies may generate spurious signal variations that impede the accurate interpretation of TL-GPR data when assessing subsurface hydrological processes. In responding to this issue, this study introduces a TL-GPR image alignment method by employing the dynamic time warping (DTW) algorithm. The purpose of the proposed method, namely TLIAM–DTW, is to correct for geometric mismatch in TL-GPR images collected from the identical survey line in the field. We validated the efficacy of the TLIAM–DTW method using both synthetic data from gprMax V3.0 simulations and actual field data collected from a hilly, forested area post-infiltration experiment. Analyses of the aligned TL-GPR images revealed that the TLIAM–DTW method effectively eliminates the influence of geometric mismatch while preserving the integrity of signal variations due to actual subsurface hydrological processes. Quantitative assessments of the proposed methods, measured by mean absolute error (MAE) and root mean square error (RMSE), showed significant improvements. After performing the TLIAM–DTW method, the MAE and RMSE between processed TL-GPR images and background images were reduced by 96% and 78%, respectively, in simple simulation scenarios; in more complex simulations, MAE declined by 27–31% and RMSE by 17–43%. Field data yielded reductions in MAE and RMSE of >82% and 69%, respectively. With these substantial improvements, the processed TL-GPR images successfully depict the spatial and temporal transitions associated with subsurface lateral flows, thereby enhancing the accuracy of monitoring subsurface hydrological processes under field conditions.

Public interest trends for COVID-19 and pandemic trajectory: A time-series analysis of US state-level data.

Google Trends provides spatiotemporal data for user-specific terms scaled from less than 1 (lowest relative popularity) to 100 (highest relative popularity) as a proxy for the public interest. Here we use US state-level data for COVID-19 to examine popularity trends during the pandemic evolution. We used "coronavirus" and "covid" search terms and set the period up from January 1st, 2020, to November 12, 2022. We measured the agreement on web rankings between states using the nonparametric Kendall's W (0 for no concordance to 1 for perfect agreement). We compiled state-level weekly data on COVID-19 incidence and mortality and scaled state curves from 0 to 100 through a min-max normalization process. We used a dynamic time-warping algorithm to calculate similarities between the popularity, mortality, and incidence of COVID-19. The methodology is a pattern recognition process between time series by distance optimization. The similarity was mapped from 0 to 1, with 1 indicating perfect similarity and 0 indicating no similarity. The peak in popularity was in March 2020, succeeded by a decline and a prolonged period of fluctuation around 20%. Public interest rose briefly at the end of 2021, to fall to a low activity of around 10%. This pattern was remarkably consistent across states (Kendal's W 0.94, p < 0.001). Web search trends were an impression of contagion growth: Overall, popularity-mortality trajectories yielded higher similarity indices (median 0.78; interquartile range 0.75-0.82) compared to popularity-incidence trajectories (median 0.74; interquartile range 0.72-0.76, Wilcoxon's exact p<0.001). The popularity-mortality trajectories had a very strong similarity (>0.80) in 19/51 (37%) regions, as opposed to only 4/51 (8%) for popularity-incidence trajectories. State-level data show a fading public concern about COVID-19, and web-search popularity patterns may reflect the COVID-19 trajectory in terms of cases and mortality.

Improving the accuracy of wind speed spatial interpolation: A pre-processing algorithm for wind speed dynamic time warping interpolation

Wind power is one of the most vital renewable energy resources in the world. Wind energy production is directly correlated with the quality and quantity of wind speed data. Interpolation techniques can be employed to fill in the gaps in the current wind speed observation data series. However, existing methods for obtaining blank data do not pre-regulate the regional spatial wind speed sequence and instead rely on direct interpolation, which leads to low accuracy in the interpolation. This is not a problem with the model itself, but rather with the wind speed flowing in space and exhibiting sequence misalignment on the timeline. To address this issue, this study proposes a Wind Speed Dynamic Time Warping (WSDTW) algorithm based on Dynamic Time Warping (DTW) to match similar wind speed reduction sequences in terms of time error. We used the shape context descriptor to encode wind speed and introduced wind rose descriptors to represent wind direction initially. The matching cost of DTW was then optimized. Finally, five common interpolation methods were selected to evaluate the method. The research results indicate that interpolation after WSDTW matching and warping can significantly improve the accuracy of wind speed interpolation and reduce the spatial dependence of wind speed. This method demonstrates good stability and generalization, performing exceptionally well in situations where there are gaps in regional wind speed data or missing data.

Data-driven classification of ligand unbinding pathways

Studying the pathways of ligand-receptor binding is essential to understand the mechanism of target recognition by small molecules. The binding free energy and kinetics of protein-ligand complexes can be computed using molecular dynamics (MD) simulations, often in quantitative agreement with experiments. However, only a qualitative picture of the ligand binding/unbinding paths can be obtained through a conventional analysis of the MD trajectories. Besides, the higher degree of manual effort involved in analyzing pathways limits its applicability in large-scale drug discovery. Here, we address this limitation by introducing an automated approach for analyzing molecular transition paths with a particular focus on protein-ligand dissociation. Our method is based on the dynamic time-warping algorithm, originally designed for speech recognition. We accurately classified molecular trajectories using a very generic descriptor set of contacts or distances. Our approach outperforms manual classification by distinguishing between parallel dissociation channels, within the pathways identified by visual inspection. Most notably, we could compute exit-path-specific ligand-dissociation kinetics. The unbinding timescale along the fastest path agrees with the experimental residence time, providing a physical interpretation to our entirely data-driven protocol. In combination with appropriate enhanced sampling algorithms, this technique can be used for the initial exploration of ligand-dissociation pathways as well as for calculating path-specific thermodynamic and kinetic properties.

A novel approach for residential load identification based on dynamic time warping

This paper proposes a new approach for load identification using a combination of fast Fourier transform (FFT) and dynamic time warping (DTW) techniques. In this approach, first, the current signals of each load device are measured individually and in various combinations to generate reference signals. FFT is then used to obtain meaningful features from the measured signals, while DTW is employed to classify the loads. This is the first known study to use the DTW algorithm for residential load identification in this structure. The proposed approach is constructed in two ways: supervised (DTW-SUP) and semi-supervised (DTW-SEM). In DTW-SEM, the reference signals for the load combinations are generated by summing the sample measurements of the individual loads, while in DTW-SUP, the actual sample measurements are used. The approach has been tested on six loads, including different load combinations. It has achieved high accuracy levels of over 90% in identifying single loads and load combinations using the error metrics Precision, Recall, and F-score. The proposed approach can potentially facilitate the implementation of energy-saving strategies in residential buildings and improve energy efficiency.

First-break prediction in 3-D land seismic data using the dynamic time warping algorithm

SUMMARY This paper presents a new methodology to assist geophysicists in determining the first-break event in a 3-D seismic data set using the well-known technique called dynamic time warping algorithm (DTW), which is usually used to find the optimal alignment between two time-series. We used the optimal path from the cost matrix to identify the first break in the seismogram using a few picks (seeds) made by an interpreter as a reference to perform this task. Furthermore, the data were pre-conditioned by the topographic and linear moveout to improve the method’s accuracy. To demonstrate the technique’s robustness, first, we applied the methodology in a synthetic seismic data. After demonstrating the efficiency of the algorithm, we applied the aforementioned methodology in the Polo-Miranga 3-D seismic cube located in the Recôncavo sedimentary basin, Bahia-Brazil, and in the seismic data acquired from the Blackfoot field in Alberta, Canada. The high-quality results showed consistency in determining the first break in all ranges of offsets, demonstrating an alternative way to accelerate this seismic processing step. Furthermore, we compared the results obtained by the proposed methodology with an algorithm based on comparing the short-time averages with long-time averages. Finally, we performed the static correction calculation to ensure that the time distortion resulting from the terrain and the low-velocity layer was mitigated in shoot gathers and in the stacked section.

Study on Rock Mechanics Parameter Prediction Method Based on DTW Similarity and Machine-Learning Algorithms

Rock mechanics parameters are crucial factors for predicting rock behavior in oil and gas reservoirs, optimizing extraction strategies, and ensuring drilling safety. In this study, we propose a random forest (RF)-convolutional neural network (CNN)-long-term short-term memory network (LSTM) fusion model based on the dynamic time warping (DTW) algorithm to construct intelligent prediction models for elastic modulus, Poisson’s ratio, and compressive strength using real-time drilling engineering data. An autoencoder with a sliding window is employed to automatically identify abnormal points or segments in the calculated values of elastic modulus, Poisson’s ratio, and compressive strength obtained from drilled wells. These abnormal values are then corrected using a backpropagation (BP) neural network. Compared to single CNN-LSTM or single RF models, the RF-CNN-LSTM fusion model performs better. It achieves this by effectively combining the strengths of different algorithms in predicting outcomes. The accuracy of the RF-CNN-LSTM fusion model is over 94% when compared to the actual values. Furthermore, the analysis of the relative importance of input parameters reveals that weight on bit (WOB), temperature, displacement, equivalent circulation density (ECD), and mud density are the primary input features for predicting elastic modulus. For predicting Poisson’s ratio, the main input features include WOB, mud density, ECD, temperature, pumping pressure, displacement, and rate of penetration (ROP). Similarly, for predicting compressive strength, the main input features consist of WOB, temperature, displacement, ECD, and mud density. The research findings demonstrate that the rock mechanics parameter prediction models based on the RF-CNN-LSTM algorithm using DTW exhibit high computational accuracy in the B oil field of China. These results are significant for gaining a deeper understanding of the variations in rock mechanics parameters and optimizing drilling decisions.

EPIC: Emotion Perception by Spatio-Temporal Interaction Context of Gait.

Recently, psychophysiological computing has received considerable attention. Due to easy acquisition at a distance and less conscious initiation, gait-based emotion recognition is considered as a valuable research branch in the field of psychophysiological computing. However, most existing methods rarely explore the spatio-temporal context of gait, which limits the ability to capture the higher-order relationship between emotion and gait. In this paper, we utilize a range of research, including psychophysiological computing and artificial intelligence, to propose an integrated emotion perception framework called EPIC, which can find novel joint topology and generate thousands of synthetic gaits by spatio-temporal interaction context. First, we analyze the joint coupling among non-adjacent joints by calculating Phase Lag Index (PLI), which can discover the latent connection among body joints. Second, to synthesize more sophisticated and accurate gait sequences, we explore the effect of spatio-temporal constraints, and propose a new loss function that utilizes the Dynamic Time Warping (DTW) algorithm and pseudo-velocity curve to constrain the output of Gated Recurrent Units (GRU). Finally, Spatial Temporal Graph Convolution Networks (ST-GCN) is used to classify emotions using the generation and the real data. Experimental results demonstrate our approach achieves the accuracy of 89.66%, and outperforms the state-of-the-art methods on Emotion-Gait dataset.

Pemanfaatan Mediapipe Body Pose Estimation dan Dynamic Time Warping untuk Pembelajaran Tari Remo

Video can be used as a learning medium for various purposes. In this research, the object of study is the movements of the traditional dance "remo." Thus, as a substitute for an absent coach or instructor, videos can take on the role of a dance instructor. However, video communication is one-way between the coach and the learners. Without movement correction, individuals trying to learn remo dance may find it challenging to determine if they are performing the movements correctly. Therefore, the author aims to create a system to assist coaches in correcting the dance movements of their learners. Using the MediaPipe module and the Dynamic Time Warping algorithm, the author developed a system to correct the learners' dance movements. This system can detect deviations from the coach's instructional video and provide notifications about which body angles do not match the coach's video instructions. The system operates by having users upload a video of their dance movements, and then it identifies which movements deviate from the correct remo dance. The accuracy is measured by comparing the angle distances between the master's movements and the test data. If the angle exceeds a predetermined threshold, the movement is considered incorrect. The system's output is validated by the coach, and it achieves 90% accuracy in identifying movement errors in videos. With this accuracy, the system can assist coaches in evaluating their learners' remo dance movements.