Activity Recognition Model Research Articles

FLSys: Toward an Open Ecosystem for Federated Learning Mobile Apps

This article presents the design, implementation, and evaluation of <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">FLSys</b> , a mobile-cloud federated learning (FL) system, which can be a key component for an open ecosystem of FL models and apps. FLSys is designed to work on smart phones with mobile sensing data. It balances model performance with resource consumption, tolerates communication failures, and achieves scalability. In FLSys, different DL models with different FL aggregation methods can be trained and accessed concurrently by different apps. Furthermore, FLSys provides advanced privacy preserving mechanisms and a common API for third-party app developers to access FL models. FLSys adopts a modular design and is implemented in Android and AWS cloud. We co-designed FLSys with a human activity recognition (HAR) model. HAR sensing data was collected in the wild from 100+ college students during a 4-month period. We implemented HAR-Wild, a CNN model tailored to mobile devices, with a data augmentation mechanism to mitigate the problem of non-Independent and Identically Distributed data. A sentiment analysis model is also used to demonstrate that FLSys effectively supports concurrent models. This article reports our experience and lessons learned from conducting extensive experiments using simulations, Android/Linux emulations, and Android phones that demonstrate FLSys achieves good model utility and practical system performance.

Research on optimization of table tennis hitting action based on image recognition technology

Abstract In the process of table tennis batting action analysis, the analysis system that relies on modern digital image processing technology plays an increasingly important role. The new GoogleNet model, as a kind of auxiliary means, because of its intuitive, rapid, and other characteristics, can be very good to find out and optimize the drawbacks of batting action in the process of the game and help learners to master the batting essentials quickly. In this paper, the batch regression algorithm is used to transform the images of players’ batting actions during table tennis games, and the AM-Softmax algorithm in the Softmax classifier is used to construct the New GoogleNet model to build the batting action recognition and analysis model that contains both temporal and spatial flows. The main conclusions are as follows: the accuracy of the test results for forehand and left-handed ball attacks based on the New GoogleNet model is as high as 92% and 90%. Forehand and left-handed ball rolling accuracy was 87.5% and 85%, respectively. The optical flow optimization method resulted in a 0.4% and 1.4% increase in the accuracy of the experiments. In the two-stream fusion method with thresholds s=99% and s=1, the accuracy of optimization of the hitting action reached 89.8% and 91.4%, respectively. The accuracy in the averaging method was 95.9% when the optical flow threshold s=1. When the optical flow threshold s = 99% is used in the downscaling method, the accuracy is 93.5%. The results of this paper are of great significance for the recognition of batting movements during table tennis matches and the optimization and improvement of batting movements.

Training and testing an artificial intelligence model for action recognition using the MMAction2 toolkit

In recent decades, artificial intelligence (AI) has become an indispensable tool in various fields of human activity, including action recognition. With the advancements in machine learning and deep learning technologies, AI's capabilities in analyzing and interpreting actions performed by humans or objects in video and audio recordings have significantly increased. This progress has led to the development of numerous applications such as surveillance systems, human-computer interaction, sports analytics, and autonomous driving, where understanding and recognizing actions is crucial. Traditional methods of action recognition relied heavily on handcrafted features and classical machine learning algorithms, which often struggled with the variability and complexity of real-world scenarios. The emergence of deep learning, particularly convolutional neural networks (CNNs) and recurrent neural networks (RNNs), has revolutionized this field by enabling models to learn hierarchical representations directly from raw data, thus improving performance and robustness. In this paper, we describe the process of training and testing an artificial intelligence model in an action recognition task using the MMAction2 toolkit.

The Effects of Different Image Parameters on Human Action Recognition Models Trained on Real and Synthetic Image Data

The Effects of Different Image Parameters on Human Action Recognition Models Trained on Real and Synthetic Image Data

Tackling Event-Based Lip-Reading by Exploring Multigrained Spatiotemporal Clues.

Automatic lip-reading (ALR) is the task of recognizing words based on visual information obtained from the speaker's lip movements. In this study, we introduce event cameras, a novel type of sensing device, for ALR. Event cameras offer both technical and application advantages over conventional cameras for ALR due to their higher temporal resolution, less redundant visual information, and lower power consumption. To recognize words from the event data, we propose a novel multigrained spatiotemporal features learning framework, which is capable of perceiving fine-grained spatiotemporal features from microsecond time-resolved event data. Specifically, we first convert the event data into event frames of multiple temporal resolutions to avoid losing too much visual information at the event representation stage. Then, they are fed into a multibranch subnetwork where the branch operating on low-rate frames can perceive spatially complete but temporally coarse features, while the branch operating on high frame rate can perceive spatially coarse but temporally fine features. Thus, fine-grained spatial and temporal features can be simultaneously learned by integrating the features perceived by different branches. Furthermore, to model the temporal relationships in the event stream, we design a temporal aggregation subnetwork to aggregate the features perceived by the multibranch subnetwork. In addition, we collect two event-based lip-reading datasets (DVS-Lip and DVS-LRW100) for the study of the event-based lip-reading task. Experimental results demonstrate the superiority of the proposed model over the state-of-the-art event-based action recognition models and video-based lip-reading models.

A biomechanics-based study of stroke technique in student tennis players

Abstract Studying the hitting technique of student tennis players is of great significance in improving the technical and tactical level of student tennis and promoting the popularization and scientificization of tennis. Since the performance of tennis hitting action recognition depends on the representation of the action, this study utilizes the Gaussian mixture distribution as a statistical distribution of human action data to provide a more accurate description of tennis hitting action. This paper presents the LM skeletal point coordinate fitting method to enhance the structure of the Gaussian mixture distribution-based Hidden Markov Model to capture the act of hitting the ball. In the end, the biomechanical properties of the tennis hitting technique were examined by analyzing the motion images of forehand and backhand hitting techniques of college tennis players. The mean values of the angle between the racket surface and the ground during the forehand stroke and the topspin stroke were 82.96° and 35.95°, respectively, and the speed of the grip point and wrist joint was faster than that of the flat stroke during the topspin stroke. The angles and velocities of the knee, hip, and ankle joints of the lower limbs were faster than those of the open backhand stroke in the closed backhand stroke. The angle and speed are smaller than that of the open backhand stroke, while the angle of the shoulder and elbow joints is larger than that of the open backhand stroke. The tennis action recognition model and algorithm designed in this paper can effectively acquire and analyze the technical characteristics of tennis hitting.

Language-assisted deep learning for autistic behaviors recognition

Correctly recognizing the behaviors of children with Autism Spectrum Disorder (ASD) is of vital importance for the diagnosis of Autism and timely early intervention. However, the observation and recording during the treatment from the parents of autistic children may not be accurate and objective. In such cases, automatic recognition systems based on computer vision and machine learning (in particular deep learning) technology can alleviate this issue to a large extent. Existing human action recognition models can now achieve impressive performance on challenging activity datasets, e.g., daily activity, and sports activity. However, problem behaviors in children with ASD are very different from these general activities, and recognizing these problem behaviors via computer vision is less studied. In this paper, we first evaluate a strong baseline for action recognition, i.e., Video Swin Transformer, on two autism behaviors datasets (SSBD and ESBD) and show that it can achieve high accuracy and outperform the previous methods by a large margin, demonstrating the feasibility of vision-based problem behaviors recognition. Moreover, we propose language-assisted training to further enhance the action recognition performance. Specifically, we develop a two-branch multimodal deep learning framework by incorporating the ”freely available” language description for each type of problem behavior. Experimental results demonstrate that incorporating additional language supervision can bring an obvious performance boost for the autism problem behaviors recognition task as compared to using the video information only (i.e., 3.49% improvement on ESBD and 1.46% on SSBD). Our code and model will be publicly available for reproducing the results.

Hybrid Convolutional Neural Network Approaches for Recognizing Collaborative Actions in Human–Robot Assembly Tasks

In the context of sustainable manufacturing, efficient collaboration between humans and machines is crucial for improving assembly quality and efficiency. However, traditional methods for action recognition and human–robot collaborative assembly often face challenges such as low efficiency, low accuracy, and poor robustness. To solve such problems, this paper proposes an assembly action-recognition method based on a hybrid convolutional neural network. Firstly, an assembly action-recognition model is proposed using skeletal sequences and a hybrid convolutional neural network model combining Spatial Temporal Graph Convolutional Networks (ST-GCNs) and One-Dimensional Convolutional Neural Networks (1DCNNs) to sense and recognize human behavior actions during the assembly process. This model combines the joint spatial relationship and temporal information extraction ability of the ST-GCN model with the temporal feature extraction ability of the 1DCNN model. By incorporating Batch Normalization (BN) layers and Dropout layers, the generalization performance of the model is enhanced. Secondly, the model is validated on a self-constructed dataset of assembly actions, and the results show that the recognition accuracy of the model can reach 91.7%, demonstrating its superiority. Finally, a digital workshop application system based on digital twins is developed. To test the effectiveness of the proposed method, three sets of control experiments were designed to evaluate both objective and subjective aspects and verify the feasibility of the method presented in this paper. Compared with traditional assembly systems, the proposed method optimizes the recognition of human–robot collaborative assembly actions and applies them to intelligent control systems using digital-twin technology. This intelligent assembly method improves assembly efficiency and saves assembly time. It enables efficient and sustainable collaboration between humans and robots in assembly, leading to a positive and sustainable impact on the manufacturing industry.

A light-weight skeleton human action recognition model with knowledge distillation for edge intelligent surveillance applications

Skeleton based human action recognition has evolved as one of the most important applications in multimedia IoT system. However, it requires extensive computation resource including high performance computing unites and large memory to train a deep mode with large number of parameters, which seriously limits it effectiveness and efficiency for edge intelligence multimedia IoT applications. In this paper, a knowledge distillation based light-weight deep model is proposed for skeleton human action recognition to meet the edge multimedia IoT applications. It can get competitive recognition performance in terms of learning accuracy for combination of AI model and edge surveillance equipment. On the one hand, to achieve desirable accuracy, we propose a deep pose-transition image representation method based on two-stream spatial–temporal architecture, which can mine the hidden features of color texture images in spatial and temporal domain, and fuse them for comprehensive discrimination before final classification. On the other hand, to increase the transfer learning ability to the student model on the edge device, we use tucker decomposition to weak the teacher model during knowledge transfer learning process. Finally, in order to validate the effectiveness of our proposal, we conducted extensive experiments to evaluate the proposed approach. The experimental results demonstrate that our proposal can realize deep model miniaturization to meet the requirement of edge multimedia IoT system and achieve the competitive performance.

Optimization-Free Test-Time Adaptation for Cross-Person Activity Recognition

Human Activity Recognition (HAR) models often suffer from performance degradation in real-world applications due to distribution shifts in activity patterns across individuals. Test-Time Adaptation (TTA) is an emerging learning paradigm that aims to utilize the test stream to adjust predictions in real-time inference, which has not been explored in HAR before. However, the high computational cost of optimization-based TTA algorithms makes it intractable to run on resource-constrained edge devices. In this paper, we propose an Optimization-Free Test-Time Adaptation (OFTTA) framework for sensor-based HAR. OFTTA adjusts the feature extractor and linear classifier simultaneously in an optimization-free manner. For the feature extractor, we propose Exponential Decay Test-time Normalization (EDTN) to replace the conventional batch normalization (CBN) layers. EDTN combines CBN and Test-time batch Normalization (TBN) to extract reliable features against domain shifts with TBN's influence decreasing exponentially in deeper layers. For the classifier, we adjust the prediction by computing the distance between the feature and the prototype, which is calculated by a maintained support set. In addition, the update of the support set is based on the pseudo label, which can benefit from reliable features extracted by EDTN. Extensive experiments on three public cross-person HAR datasets and two different TTA settings demonstrate that OFTTA outperforms the state-of-the-art TTA approaches in both classification performance and computational efficiency. Finally, we verify the superiority of our proposed OFTTA on edge devices, indicating possible deployment in real applications. Our code is available at https://github.com/Claydon-Wang/OFTTA.

Using complex networks and multiple artificial intelligence algorithms for table tennis match action recognition and technical-tactical analysis

This study developed two action recognition models using the YOLOv8-Alphapose two-stream spatial temporal graph convolutional networks (2s-STGCN), and the networks were used to recognize technical actions in table tennis. This study proposed a novel framework that merges dynamic and static complex network analysis with a community detection algorithm aimed at evaluating table tennis players' techniques, tactical patterns and styles. Two datasets that contain 8015 high-definition action videos of 37 elite players were constructed: a front-facing player technical action dataset (4154 videos) and a backwards-facing player technical action dataset (3861 videos). The results showed that YOLOv8-Alphapose-2s-STGCN achieved better recognition performance than seven other YOLOv8-Alphapose-based artificial intelligence algorithms (transformer, BiGRU, BiLSTM, GRU, LSTM, TCN and RNN algorithms) on both datasets and exhibited robust performance in practical applications. In the case study, multiple indicators were used to measure the importance of nodes (players' techniques) within the serving and receiving networks and within the two-round (winning and losing) networks. Dynamic complex network analysis was adopted to evaluate tactical styles and patterns. Furthermore, this study examined whether players and their opponents exhibit variability or similarity in their tactical patterns, focusing on the player networks and the two-round winning and losing networks. By integrating action recognition with process-focused match analysis, this study explored an innovative and comprehensive way to analyse matches, with implications for the performance analysis of table tennis players and players in related racket sports.

Seq2seq model for human action recognition based on skeleton and two-layer bidirectional LSTM

Human action recognition (HAR) plays an important role in social interaction in various fields. This study proposes a light-weight skeleton and two-layer bidirectional LSTM-based Seq2Seq model (SB2_Seq2Seq) for HAR to trade off recognition accuracy, users’ privacy and computer resource usage. An experiment was conducted to compare the proposed SB2_Seq2Seq with other skeleton-based Seq2Seq models and non-skeleton RGB video frame-based LSTM, CNN and seq2seq models. The UCF50 dataset was used for model evaluation, where 60%, 20% and 20% for model training, validation and testing, respectively. The experimental results show that the proposed model achieves 93.54% accuracy with 0.0214 Mean Square Error (MSE), suggesting that the proposed model outperforms all the other models. Besides, it also shows that the proposed model achieves state-of-the-art accuracy compared with state-of-the-arts methods in literature.

STMultiple: Sparse Transformer Based on RFID for Multi-Object Activity Recognition

Wireless sensing techniques for Human Activity Recognition (HAR) have been widely studied in recent years. At present, research on HAR based on Radio Frequency Identification (RFID) is changing from the tag attachment method to the tag non-attachment method. Affected by multipath, the current solutions in tag non-attachment scenarios mainly focus on single-object activity recognition, which is not suitable for multi-object scenarios. To address these issues, we propose STMultiple, a novel tag non-attachment activity recognition model for multi-object. The model first preprocesses the raw signal with filter and phase calibration, then it applies dilated convolution in the frequency domain to extract multi-object activity features, finally the feature pyramid structure and ProbSparse are used to optimize the vanilla Transformer-Encoder to enhance the activity recognition ability. Extensive experiments show that STMultiple can achieve recognition accuracy of up to 97.93% and down to about 90% in challenging environments ranging from two to five users, which has excellent performance compared to several state-of-the-art methods.

Action Recognition in Dark Videos Using Spatio-Temporal Features and Bidirectional Encoder Representations From Transformers

Several research works have been developed in the area of action recognition. Unfortunately, when these algorithms are applied to low-light or dark videos, their performances are highly affected and found to be very poor or fall rapidly. To address the issue of improving the performance of action recognition in dark or low-light videos; in this article, we have developed an efficient deep 3D CNN based action recognition model. The proposed algorithm follows two-stages for action recognition. In the first stage, the low-light videos are enhanced using Zero-Reference Deep Curve Estimation (Zero-DCE), followed by the min-max sampling algorithm. In the latter stage, we propose an action classification network to recognize the actions in the enhanced videos. In the proposed action classification network, we explored the capabilities of the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$R(2+1)D$</tex-math></inline-formula> for spatio-temporal feature extraction. The model's overall generalization performance depends on how well it can capture long-range temporal structure in videos, which is essential for action recognition. So we have used a Graph convolutional network (GCN) on the top of R(2+1)D as our video feature encoder which captures long-term temporal dependencies of the extracted features. Finally, a Bidirectional Encoder Representations from Transformers (BERT) is adhered to classify the actions from the 3D features extracted from the enhanced video scenes. The effectiveness of the proposed action recognition scheme is verified on ARID V1.0 and ARID V1.5 datasets. It is observed that the proposed algorithm is able to achieve <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$96.60\%$</tex-math></inline-formula> and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$99.88\%$</tex-math></inline-formula> as Top-1 and Top-5 accuracy, respectively, on ARID V1.0 dataset. Similarly, on ARID V1.5, the proposed algorithm is able to achieve <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$86.93\%$</tex-math></inline-formula> and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$99.35\%$</tex-math></inline-formula> as Top-1 and Top-5 accuracies, respectively. To corroborate our findings, we have compared the results obtained by the proposed scheme with those of fifteen state-of-the-art action recognition techniques.

Human Movement Recognition Based on 3D Point Cloud Spatiotemporal Information from Millimeter-Wave Radar

Human movement recognition is the use of perceptual technology to collect some of the limb or body movements presented. This practice involves the use of wireless signals, processing, and classification to identify some of the regular movements of the human body. It has a wide range of application prospects, including in intelligent pensions, remote health monitoring, and child supervision. Among the traditional human movement recognition methods, the widely used ones are video image-based recognition technology and Wi-Fi-based recognition technology. However, in some dim and imperfect weather environments, it is not easy to maintain a high performance and recognition rate for human movement recognition using video images. There is the problem of a low recognition degree for Wi-Fi recognition of human movement in the case of a complex environment. Most of the previous research on human movement recognition is based on LiDAR perception technology. LiDAR scanning using a three-dimensional static point cloud can only present the point cloud characteristics of static objects; it struggles to reflect all the characteristics of moving objects. In addition, due to its consideration of privacy and security issues, the dynamic millimeter-wave radar point cloud used in the previous study on the existing problems of human body movement recognition performance is better, with the recognition of human movement characteristics in non-line-of-sight situations as well as better protection of people’s privacy. In this paper, we propose a human motion feature recognition system (PNHM) based on spatiotemporal information of the 3D point cloud of millimeter-wave radar, design a neural network based on the network PointNet++ in order to effectively recognize human motion features, and study four human motions based on the threshold method. The data set of the four movements of the human body at two angles in two experimental environments was constructed. This paper compares four standard mainstream 3D point cloud human action recognition models for the system. The experimental results show that the recognition accuracy of the human body’s when walking upright can reach 94%, the recognition accuracy when moving from squatting to standing can reach 84%, that when moving from standing to sitting can reach 87%, and the recognition accuracy of falling can reach 93%.

Deep Convolutional Neural Network with Symbiotic Organism Search-Based Human Activity Recognition for Cognitive Health Assessment.

Cognitive assessment plays a vital role in clinical care and research fields related to cognitive aging and cognitive health. Lately, researchers have worked towards providing resolutions to measure individual cognitive health; however, it is still difficult to use those resolutions from the real world, and therefore using deep neural networks to evaluate cognitive health is becoming a hot research topic. Deep learning and human activity recognition are two domains that have received attention for the past few years. The former is for its relevance in application fields like health monitoring or ambient assisted living, and the latter is due to their excellent performance and recent achievements in various fields of application, namely, speech and image recognition. This research develops a novel Symbiotic Organism Search with a Deep Convolutional Neural Network-based Human Activity Recognition (SOSDCNN-HAR) model for Cognitive Health Assessment. The goal of the SOSDCNN-HAR model is to recognize human activities in an end-to-end way. For the noise elimination process, the presented SOSDCNN-HAR model involves the Wiener filtering (WF) technique. In addition, the presented SOSDCNN-HAR model follows a RetinaNet-based feature extractor for automated extraction of features. Moreover, the SOS procedure is exploited as a hyperparameter optimizing tool to enhance recognition efficiency. Furthermore, a gated recurrent unit (GRU) prototype can be employed as a categorizer to allot proper class labels. The performance validation of the SOSDCNN-HAR prototype is examined using a set of benchmark datasets. A far-reaching experimental examination reported the betterment of the SOSDCNN-HAR prototype over current approaches with enhanced precision of 86.51% and 89.50% on Penn Action and NW-UCLA datasets, respectively.

A multimedia art interaction model for motion recognition based on data-driven model

Ergonomics is a relatively important part of user experience in multimedia art design. The study aims to improve human–computer interaction efficiency using a data-driven neural network model combined with video data and wearable devices to achieve high-precision human motion recognition. Firstly, the human motion skeleton information is extracted based on video information through the OpenPose framework, and the motion characteristics of each joint of the human body are calculated. Then, use the inertia data of the wearable bracelet to calculate the exercise intensity. At last, the two kinds of information are sent to recurrent neural networks (RNN) together to achieve high-precision human motion recognition. The innovative contribution of the article is to establish a multimodal fusion model for human activity recognition. The experimental results show that the recognition precision of the proposed method reaches 97.85%, much higher than the backpropagation neural network (BPNN) and K-nearest neighbor (KNN), whose precision is 94.35% and 90.12%, respectively. The method’s superior performance convinces us that the model can provide strong technical support for the interaction design of multimedia art.

Towards efficient video-based action recognition: context-aware memory attention network

Abstract Given the prevalence of surveillance cameras in our daily lives, human action recognition from videos holds significant practical applications. A persistent challenge in this field is to develop more efficient models capable of real-time recognition with high accuracy for widespread implementation. In this research paper, we introduce a novel human action recognition model named Context-Aware Memory Attention Network (CAMA-Net), which eliminates the need for optical flow extraction and 3D convolution which are computationally intensive. By removing these components, CAMA-Net achieves superior efficiency compared to many existing approaches in terms of computation efficiency. A pivotal component of CAMA-Net is the Context-Aware Memory Attention Module, an attention module that computes the relevance score between key-value pairs obtained from the 2D ResNet backbone. This process establishes correspondences between video frames. To validate our method, we conduct experiments on four well-known action recognition datasets: ActivityNet, Diving48, HMDB51 and UCF101. The experimental results convincingly demonstrate the effectiveness of our proposed model, surpassing the performance of existing 2D-CNN based baseline models. Article Highlights Recent human action recognition models are not yet ready for practical applications due to high computation needs. We propose a 2D CNN-based human action recognition method to reduce the computation load. The proposed method achieves competitive performance compared to most SOTA 2D CNN-based methods on public datasets.

A Survey on Different Deep Learning Model for Human Activity Recognition based on Application

The field of human activity recognition (HAR) seeks to identify and classify an individual's unique movements or activities. However, recognizing human activity from video is a challenging task that requires careful attention to individuals, their behaviors, and relevant body parts. Multimodal activity recognition systems are necessary for many applications, including video surveillance systems, human-computer interfaces, and robots that analyze human behavior. This study provides a comprehensive analysis of recent breakthroughs in human activity classification, including different approaches, methodologies, applications, and limitations. Additionally, the study identifies several challenges that require further investigation and improvements. The specifications for an ideal human activity recognition dataset are also discussed, along with a thorough examination of the publicly available human activity classification datasets.

Design and implementation of adolescent health Latin dance teaching system under artificial intelligence technology.

Since various dance teaching systems have attracted much attention with the development of Artificial Intelligence (AI) technology, this paper improves the recognition performance of Latin dance teaching systems by optimizing the action recognition model. Firstly, the object detection and action recognition technology under the current AI technology is analyzed, and the Two-stage object detection algorithm and One-stage object detection algorithm are evaluated. Secondly, the technologies and functions contained in the adolescent health Latin dance teaching system are described, including image acquisition, feature extraction, object detection, and action recognition. Finally, the action recognition algorithm is optimized based on object detection, and the rationality and feasibility of the proposed algorithm are verified by experiments. The experimental results show that the optimization algorithm can search the optimal feature subset after five iterations on Undefine Classes of 101 (UCF101) dataset, but it needs seven iterations on Human Motion Database 51 (HMDB51) dataset. Meanwhile, when using support vector machine classifier, the optimization algorithm can achieve the highest accuracy of motion recognition. Regressive Function, Multinomial Naive Bayes and Gaussian Naive Bayes Algorithms have lower prediction delay, as low as 0.01s. Therefore, this paper has certain reference significance for the design and implementation of adolescent health Latin dance teaching system.