This protocol for the review was implemented following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement [35], which has been registered in the PROSPERO (International Prospective Register of Systematic Reviews) database under the ID CRD42024539375. Literature was retrieved from the following 5 databases: PubMed, Embase, Scopus, IEEE Xplore, and ACM Digital Library. The focus of this paper is the application, research trends, and clinical efficacy of the TCC auxiliary training system. There are many spellings of Medical Subject Headings terms for TCC, including Taiji, Tai Chi, TCC, and T’ai Chi. We modified the retrieval strategy according to different databases, and the search terms of all databases are provided in Multimedia Appendix 1. This study summarizes the studies published from January 2014 to May 2024.

Studies eligible to be included in this review had to meet the following inclusion criteria: (1) studies describing a TCC training system, (2) the search language is limited to English, and (3) studies published in the past 10 years. The exclusion criteria were (1) systematic reviews or literature reviews, (2) books or study comments, (3) theses, (4) incomplete or short papers (eg, posters, tutorials, and technical reports), and (5) primary studies that are repeated.

This search included journal and conference papers. The study screening process consisted of the following steps: (1) retrieving studies from January 2014 to May 2024, (2) screening titles and abstracts of the remaining studies after removing duplicates, and (3) reviewers reading the full texts and selecting studies according to the inclusion and exclusion criteria. If a journal study covers the content reported in the previous conference paper or degree thesis, the journal paper was given precedence over the conference paper and degree thesis.

After establishing the inclusion and exclusion criteria, the next step was to identify quality criteria to strengthen the extraction of quantitative and qualitative data for the synthesis and results analysis. We established a list of 5 quality standards (Table 1) based on Santos et al [36], aligning with the objectives of this review on TCC auxiliary training systems. Each study was rated by 2 authors (HuibiaoLi and HH) according to these 5 specific criteria. Finally, the final score of each researcher was sorted by Hong Liu, and any differences were checked. Any discrepancies between the raters were resolved through negotiation. Each criterion in the list is rated as follows: yes=1 point, no=0 points, and partial=0.5 points. The total score was derived by summing the scores of the questions, reflecting the overall quality of the included literature: 0‐2 (low quality), 3‐4 (medium quality), and 5 (high quality).

After identifying eligible publications, all relevant data were collected in Microsoft Excel using a structured coding scheme. Data were extracted independently by 2 authors (Huibiao Li and HH). The collected variables included title, sample size, year, country, research methods and techniques, system description, type of virtual environment, system evaluation and validation, case study results, development tools, interaction design, and categories of TCC. Detailed descriptions were provided for development tools, system design, evaluation and validation methods, and clinical efficacy studies. Any discrepancies were resolved through discussion and re-evaluation of the relevant literature.

The data synthesis process used a narrative approach to analyze and present the findings of the included studies. An overview of the development tools used by different systems, descriptions of system designs, and evaluations or validations for each system were provided to facilitate cross-study comparisons. This study used summarized data wherever possible and followed the PRISMA-ScR (Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews) guidelines.

Figure 1 shows an overview of the study selection results at different stages. Of the 2202 studies retrieved through the search strategies, 34 papers were selected for inclusion in this review. Table 2 shows the quality of the studies included in the review. According to our predefined quality assessment criteria, 2 of the 34 studies included in the study were considered high quality, 22 were considered medium quality, and 10 were considered low quality. In summary, 94% (32/34) of the research schemes described the methods and techniques used, 82% (28/34) described the TCC auxiliary training system, 32% (11/34) reported the virtual environment used, 50% (17/34) evaluated or verified the system, and 41% (14/34) provided detailed case study results.

Table 3 shows the year, country (based on the research unit of the first author), and type of publication included in the review. In the research we included, 65% (22/34) of the studies were conference papers, and the remainder were journal papers. Since TCC is a traditional martial art originating in China, 76% (26/34) of the studies are from China.

Figure 2 shows the temporal trend of the development of the TCC auxiliary training system from 2014 to 2024. This trend shows fluctuations in research over the past decade. In terms of time trends, research on TCC auxiliary training systems has been increasing in the past 10 years, with most system development research focusing on the period before and after the outbreak of the COVID-19 pandemic (59%, 20/34).

This section summarizes the work conducted for evaluating or verifying each system according to the VR-core framework [74], a framework commonly used for evaluating VR systems, which includes evaluating system acceptability, feasibility, tolerability, and clinical efficacy.

To enhance the innovation and foresight of the TCC auxiliary training system, we further explored its future development directions and proposed several potential technology integration approaches. These are aimed at addressing current challenges in personalized training, real-time feedback, and data analysis, thereby promoting the precision and intelligence of TCC training outcomes.

The results of this review indicate that most individuals can effectively engage in TCC training using assistive systems, which offer significant benefits for both physical and mental health. Existing studies show that the development tools for TCC auxiliary training systems are diverse, enabling the creation of training environments through various hardware and software combinations. The design of interaction modes and feedback systems effectively enhances participants’ engagement, learning interest, and training outcomes, without causing significant adverse effects. However, clinical validation of TCC auxiliary training systems remains insufficient. This necessitates future research to conduct longitudinal studies with standardized reporting to strengthen clinical validation. Finally, this study offers forward-looking insights by providing practical, design-oriented recommendations, such as integrating emerging technologies. These recommendations aim to guide future development, an area that has not been sufficiently explored in previous practice. Therefore, this review not only synthesizes existing evidence but also provides a strategic roadmap for designing more scalable and feasible TCC auxiliary training systems.

Immersive VR is principally used in developing the VR environment of existing TCC auxiliary training systems. The TCC assistant training system based on immersive VR has significant advantages in improving the learning speed of TCC practitioners, among which the HMD environment with the best VR experience has the fastest learning speed. In addition, compared with personal computers and HMD, the CAVE environment is more helpful for improving learning quality and increasing action learning efficiency [32]. This may be because the VR environment provided by HMD and CAVE becomes the visual reality of the practitioner without any hint of the physical environment. By using sensory information related to exercise, practitioners improve their attention to learning and motivation and, subsequently, the learning efficiency of TCC. The virtual scene design of mixed reality is also used in the current system design. In mixed reality, TCC practitioners can coexist with virtual coaches and interact in a mixed environment. After wearing a helmet, the practitioner can see many virtual coaches around them, and the TCC movements of the coach from different angles can be easily observed. Simultaneously, movements can be corrected over time through the augmented mirror [42]. In contrast, the AR design in the practice system is closer to the real-world scene, introducing computer-generated elements, including the use of common devices such as smartphones or tablets as viewing media, thereby enhancing the sensory experience of the natural environment. Therefore, existing research [55] shows that using an AR-based TCC auxiliary training system for TCC training can effectively improve the balance control of older adults and increase the muscle strength of the lower extremities.

Microsoft HoloLens is the most widely used VR tool in implementing VR environments. It accurately tracks TCC practitioners’ head movements in real time and offers immediate interactive experiences [42,51,53]. Its enhanced features and improved health services also cater to clinical rehabilitation and medical environments [85]. Second, the VR tool Oculus Rift has the characteristics of high resolution, a large field of view, low weight, easy setup, and easy access to good driver support [44]. However, it is essential to note that if the training system is to be applied to older adults or people with dysfunction, then the weight of HMD must be considered. In addition, although the use of HMD for TCC learning has achieved good learning results, we must consider that helmet displays are limited to a certain extent in the field of vision.

Kinects are widely adopted for motion capture due to their low cost, portability, and markerless capabilities, enabling data collection in diverse environments. They are particularly prevalent in VR training systems and are recognized as safe, effective, and feasible tools for rehabilitation in geriatric, neurological, and sports settings [50,86,87]. However, the disadvantage of Kinect-based systems is lower capture accuracy. There are 2 solutions in the current system design to solve this problem. One is to use the high-precision Vicon system as the motion acquisition equipment for the TCC coach, while the practitioner uses Kinect to capture the motion. This system design ensures the professionalism and accuracy of the coach action database and increases system portability, which is no longer limited by time and place, allowing it to be used more widely. Another method is to use a multimodal input combination. For example, based on Kinect, the plantar pressure sensor works together to obtain motion data from TCC practitioners and instantiates virtual images from practitioners as real-time input, effectively increasing the interactivity of the virtual environment [44,49,58]. Notably, the validity of the kinematic data recorded by Kinect should be improved. The book “Virtual Reality for Physical and Motor Rehabilitation” recommends camera placement to obtain the best motion tracking for upper limb applications. The Kinect camera should be located within 30×30 cm², at a distance of 1.45 to 1.75 meters from the user, and 0.15 meters to either the left or right [88]. This study can provide a reference for system development and design in the future.

Another multimodal interaction uses wearable devices, Noitom and HMD, as action and voice inputs, resulting in natural voice interaction and gesture interaction, which creates a stronger sense of immersive experience and dramatically increases the practitioner’s sense of participation. Simultaneously, from the perspective of neural rehabilitation, this visual-auditory interaction can effectively improve the cognitive ability of users, especially attention and working memory [89]. With respect to current technology, achieving a complete multimodal interactive VR environment is not easy. The ultimate goal of human-computer interaction is to achieve seamless interaction as if the TCC coach and the practitioner were interacting face-to-face. Therefore, based on multimodal means including gestures, voice, and touch, we should also consider vision-based facial expression capture, mental state assessment, and wearable physiological index detection.

When the TCC auxiliary training system converts the TCC action of the practitioner into signal data that can be calculated and analyzed through multiple input devices, the algorithm is then used for segmenting, feature extraction, and classification of signal data. The most frequently mentioned classification algorithm is the DTW algorithm based on template matching. In the TCC auxiliary training system, the algorithm is mainly used for similarity matching. With the wide application of the algorithm, several researchers have innovated and improved it. A study used 8 bone vectors from human bones and body directions as input features and proposed a concise version of the DTW algorithm, which can further convert DTW distances into meaningful performance scores without requiring expert training data and experience [50]. The algorithm was effective and consistent compared to the expert score. However, the background of the 3 scoring experts was not introduced, so the results require further validation.

In recent years, machine learning has emerged as one of the forefront technologies in AI and is undergoing rapid development. Neural network models relevant to machine learning have found extensive applications across various domains, including video analysis and digital image processing. The TCC auxiliary training system has recently started using neural network models for human posture estimation. Some studies transformed the human posture estimation problem into a neural network–based skeletal joint regression problem [61,68]. Compared to conventional motion capture systems, another study proposed a TCC auxiliary training system based on pose estimation using convolutional neural networks, which demonstrated enhanced accuracy in estimating the postures of TCC practitioners [54].

The TCC auxiliary training system helps practitioners complete the required movements through continuous human-computer interactions. Therefore, after applying the algorithm to evaluate the collected motion data, it is necessary to provide feedback on motion quality to the practitioner in real time to improve the degree of participation. According to the literature, real-time visual feedback is the most widely used feedback strategy. Through real-time visual feedback, TCC practitioners can see themselves on the same screen as the instructor in the video [56], as well as the center of gravity distribution [39,49,51] and breathing status [39], which helps practitioners quickly improve their physical movements, enhance their interest in learning, and improve learning efficiency. In addition, some studies have designed 3 exaggerated real-time virtual visual cues to display the center of mass distribution. This exaggerated design enables practitioners to more easily observe the center of mass distribution of the virtual coach, providing useful visual feedback, particularly suited for beginners [51]. Simultaneously, exaggerated virtual sports design is more entertaining, increasing practitioner engagement in the practice process and improving virtual interaction performance. Introducing virtual smoke into the TCC training system to enhance the effect of motion display can encourage practitioners to practice longer, support sports memory, and aid memory retention [52,90].

With the advancement of technology, multimodal interaction has provided TCC practitioners with a more enriched and personalized learning experience. The feedback-based spatial path teaching method helps learners achieve better training outcomes through feedback and guidance. In addition, to emphasize training personalization, effects such as voice prompts, track settings, and arrow indicators are introduced for learners to choose, which improves training effectiveness and user participation [65]. To overcome the interactive limitations of traditional teaching methods and enable TCC learners to fully experience the joy of motion learning through natural interactions with the system, multimodal interaction via multiplatform browsers has been used to seamlessly and swiftly interact with the system, controlling the scenes and characters in the web and achieving effective learning objectives [67].

Notably, some researchers [91] believe that real-time feedback may be more effective for beginners, while terminal feedback is more suitable for skilled users. Therefore, combining real-time and terminal feedback can be considered in future system design, stimulating practitioners’ efforts, motivation, and persistence to some extent and potentially exceeding the goals achieved in current practice.

In our opinion, the TCC auxiliary training system based on VR serves both as an auxiliary training system and a medical tool for rehabilitation exercises. However, the use of this system in clinical settings must be clinically validated. Of the studies we reviewed, only 2 studies clinically validated the designed system. Compared to traditional coach-guided training groups, the AR-based TCC auxiliary training system is effective for rehabilitation in older adults, and its clinical efficacy has been validated [55]. In addition, combining the VR-based TCC auxiliary training system can offer VR and TCC exercise programs for older adults with cognitive impairments. Compared to the control group maintaining regular physical activity, this system preserves cognitive and physical functions in older adults with cognitive impairments [48].

Although the results mentioned in the above "Clinical Efficacy" section are promising, these systems still lack robust research designs, rigorous measurement methods, and standardized, appropriate clinical outcome measures. This not only affects the external validity of the intervention effects but also limits the implementation of large-scale, multicenter randomized controlled trials to some extent. Particularly, in the context of rapid system iterations, the stability and generalizability of these systems in long-term clinical pathways remain unclear.

To advance the field, we recommend that future studies: first, adopt multicenter, randomized controlled trials with sufficient sample sizes, focusing on populations with high clinical value, such as older adults at high risk of falls, patients with Parkinson disease, and those with mild cognitive impairment. The control group may receive traditional rehabilitation training, health education, or VR experiences with placebo effects, ensuring clear attribution of the intervention effects. The intervention period should be at least 8 weeks, with a follow-up period of 3 to 6 months, and measurements taken at multiple time points to assess the maintenance effects of the intervention. Second, future studies should clearly define prioritized outcome measures in the research protocol and align them with the 4 dimensions of the VR core framework: acceptability, feasibility, tolerance, and clinical efficacy. For example, clinical trials targeting older adults may include primary outcomes such as balance or cognitive function, and secondary outcomes such as quality of life, lower limb strength, fall incidence, and the 4 dimensions of the VR core framework. Finally, the TCC system should be integrated into clinical rehabilitation pathways or telemedicine frameworks. It could be implemented in rehabilitation departments of tertiary hospitals, where the TCC system can serve as an adjunct module for physical therapy or occupational therapy in inpatient and outpatient rehabilitation processes. In community rehabilitation centers, TCC could be promoted as a group or individualized intervention to enhance its accessibility. In addition, remote medical platforms could integrate head-mounted displays, motion sensors, and cloud-based monitoring to enable real-time guidance and data feedback in home settings, allowing physicians and rehabilitation therapists to remotely assess progress and adjust training programs [92].

Despite the VR core framework recommending a comprehensive evaluation of VR-based interventions across 4 key domains (acceptability, feasibility, tolerance, and clinical efficacy), most of the included studies in this review did not report all these dimensions. This omission limits the ability to fully assess the value and applicability of the TCC auxiliary training system. The lack of acceptability and feasibility data hinders understanding of whether the intervention can be practically adopted in various settings, while the absence of tolerance data impedes a robust assessment of safety and user comfort. Without this information, even positive clinical efficacy results may overestimate the practical applicability of these systems. Potential reasons for this reporting gap include the early development stages of many systems, heterogeneous study designs, and a focus in most studies on efficacy outcomes rather than comprehensive evaluations. Future research should incorporate all 4 domains of the VR core framework into study protocols and report using standardized measures, leading to more balanced and generalizable conclusions regarding the efficacy and implementation potential of VR-assisted TCC training systems, with higher comparability across studies.

In recent years, there have been numerous new findings regarding the clinical efficacy of TCC, with traditional TCC practices proving highly effective in reducing stress and promoting meditation or relaxation. The research findings of Kim et al [64] also demonstrated that virtual TCC practices can induce relaxation, reduce stress, and promote mental tranquility. This suggests that the virtual TCC auxiliary training system can serve as a mindfulness and meditation tool for promoting emotional well-being in older adults. In addition, TCC auxiliary training systems enable individuals to engage effortlessly and authentically in TCC practices, significantly reducing physical and psychological barriers. TCC auxiliary training systems can create both standing and sitting TCC practices [44,93]. Sitting TCC can reduce the pressure on the joints to the greatest extent and eliminate pain or fear caused by exercise. Therefore, sitting during TCC practice can make the practitioner’s mind more relaxed and peaceful. Moreover, the TCC auxiliary training system offers unique features such as engaging virtual environments, soothing background music, and a scoring system. These features are attractive and motivating for participants, facilitating the execution of movements and meditation techniques.

The TCC auxiliary training systems are suitable for people of all ages, including those with chronic diseases or dysfunction. Therefore, these systems are effective for home and community rehabilitation and can maintain the continuity of neural rehabilitation by overcoming social obstacles such as distance and cost, allowing practitioners to adjust their training schedule and exercise intensity. For practitioners, this is an entertaining and healthy way of life that positively impacts the mental health, physical health, self-esteem, and attention of older adults.

Notably, only 2 of the included studies explicitly assessed and reported clinical efficacy, highlighting a critical gap in the current evidence base. Clinical efficacy is the most direct indicator of the health benefits of the TCC auxiliary training system. Although other metrics, such as acceptability or feasibility, may be favorable, the lack of clinical efficacy data limits the ability to determine the real-world therapeutic value of these interventions. This scarcity may be attributed to the early development stage of most systems, small sample sizes, short intervention durations, and a primary focus on technical validation rather than patient-centered outcomes. Therefore, future research should prioritize well-powered, well-controlled trials that include validated clinical outcome measures as primary endpoints, while also conducting comprehensive assessments of acceptability, feasibility, and tolerability to provide an overall evaluation of intervention impact.

This review has several limitations. First, our search strategy included studies broadly related to the TCC auxiliary training system but did not systematically target specific diseases or populations. As a result, this review may not fully capture TCC auxiliary training systems tailored to specific diseases (eg, depression, anxiety, bipolar disorder, or post-traumatic stress disorder) or populations (eg, middle-aged individuals, older adults, or beginners). Future research should adopt more refined approaches to evaluate TCC auxiliary training systems targeting different diseases or populations, providing a deeper understanding of their effectiveness for these groups.

Second, the lack of standardized clinical efficacy assessment criteria is a major challenge in this field. The assessment tools and clinical indicators used in different studies are often inconsistent, leading to poor comparability and consistency of the findings. Clinical efficacy is typically assessed based on short-term effects, with a lack of long-term follow-up studies, making the evaluation of the TCC auxiliary training system’s efficacy and lasting impact incomplete. This limits the credibility of broader adoption and clinical application. A standardized evaluation system should be established for TCC auxiliary training systems to improve the reliability of assessing their therapeutic effects.

Third, this review found that most of the included studies were conducted in China, likely due to TCC’s origin and widespread practice in the country. This geographic concentration may limit the generalizability of our findings to populations in other cultural and health care contexts. Factors such as familiarity with TCC culture, local rehabilitation practices, and available technological infrastructure may influence user engagement and intervention outcomes. In addition, although our search was limited to English-language databases, the dominance of studies from a single country may introduce geographical bias. Future research should include multicenter trials conducted in different geographic regions to validate the applicability and effectiveness of TCC-based training systems in diverse cultural and health care settings.

Finally, while this review addresses ethical issues related to TCC auxiliary training systems to a lesser extent, empirical research examining the real-world ethical, social, and clinical impacts remains limited. The ethical impacts and accessibility challenges associated with deploying TCC auxiliary training systems are critical and cannot be overlooked. On one hand, while the application of body-tracking technology can provide accurate motion data and personalized feedback, it also raises significant concerns about privacy protection. TCC training systems use high-precision sensors and cameras to monitor practitioners’ movements, posture, and physiological states in real time, thereby increasing the risk of personal data breaches, particularly during data storage, transmission, and analysis. This issue is especially concerning older adults or vulnerable populations, who may not fully recognize the potential risks of data collection or how their data will be used, thereby complicating privacy protection. Ensuring data anonymization, legal and compliant use, and informed user consent is therefore crucial. Accordingly, the system should establish a robust data privacy protection mechanism to ensure that users can make informed and autonomous choices.

On the other hand, another accessibility challenge for TCC auxiliary training systems is their applicability to frail older adults. Frail older adults often face multiple physical, cognitive, and psychological challenges, such as poor motor coordination, slow response times, and difficulties understanding and operating complex technologies. While TCC, as a mild form of exercise, can improve the physical health of older adults, complex system operations and interfaces may impose a burden on some frail individuals. Furthermore, the precision of the personalized training system and the intensity of real-time feedback may need further adaptation and adjustment based on the physical conditions of older adults. To ensure that the TCC auxiliary training system benefits all groups, system design must account for the diverse needs of users with varying physical conditions, provide an intuitive interface, and reduce technical barriers through personalized features. This will not only help older users overcome technical adaptation challenges but also ensure the system’s sustainability and wide applicability in real-world use. Despite these limitations, this review provides a comprehensive overview of the application of TCC auxiliary training systems in the current context, identifies key research gaps, and offers feasible suggestions for future studies.

This review provides a comprehensive evaluation of the design, application, research trends, and clinical effectiveness of TCC auxiliary training systems and offers recommendations for future development. We followed the PRISMA-ScR guidelines and analyzed 34 peer-reviewed studies published after 2014. This review addresses 4 key questions (RQs): development tools (RQ1), system design (RQ2), evaluation or validation (RQ3), and future development (RQ4). It emphasizes current development trends of TCC auxiliary training systems and outlines the design framework required for future advancements.

For RQ1, our findings indicate that development tools for TCC auxiliary training systems are diverse in both hardware and software. In terms of hardware, motion capture devices, VR tools, and AR technologies are widely used. In terms of software, Unity3D plays a leading role in system development, aiding in the creation of high-quality virtual training environments. However, further attention is needed to integrate various types of motion capture technologies and develop more immersive systems to enhance training effectiveness.

Regarding RQ2, we found that the design of TCC auxiliary training systems addresses certain demographic needs, with the 24-form Tai Chi style as the primary reference, incorporating various interaction modes and algorithms. However, further optimization is needed in integrating different interaction modes and considering accessibility to enhance training outcomes and user engagement.

For RQ3, the findings suggest that TCC auxiliary training systems show promising results in acceptability, feasibility, tolerability, and clinical efficacy. However, further optimization of system design is needed to address existing usability barriers and conduct large-scale longitudinal studies to strengthen clinical validation. Finally, for RQ4, future research could further optimize the integration of technologies and promote interdisciplinary collaboration to enhance the intelligence and precision of TCC auxiliary training systems.