Toward Personalized Darts Training: A Data-Driven Framework Based on Skeleton-Based Biomechanical Analysis and Motion Modeling
Authors: Zhantao Chen, Dongyi He, Jin Fang, Xi Chen, Yishuo Liu, Xiaozhen Zhong, Xuejun Hu
Organizations: School of Computer Technology and Application, Qinghai University, Qinghai, 810016, China · Xiaoping Technology Innovation Lab, Zhongshan Xiaolan Senior High School, Guangdong, 528415, China · School of Artificial Intelligence, Chongqing University of Technology, Chongqing, 400054, China · Department of Language Science and Technology, The Hong Kong Polytechnic University, Hung Hom, 999077, Hong Kong SAR, China · Faculty of Education, University of Macau, Taipa, 999078, Macau SAR, China
As sports training becomes more data-driven, traditional dart coaching based mainly on experience and visual observation is increasingly inadequate for high-precision, goal-oriented movements. Although prior studies have highlighted the importance of release parameters, joint motion, and coordination in dart throwing, most quantitative methods still focus on local variables, single-release metrics, or static template matching. These approaches offer limited support for personalized training and often overlook useful movement variability. This paper presents a data-driven dart training assistance system. The system creates a closed-loop framework spanning motion capture, feature modeling, and personalized feedback. Dart-throwing data were collected in markerless conditions using a Kinect 2.0 depth sensor and an optical camera. Eighteen kinematic features were extracted from four biomechanical dimensions: three-link coordination, release velocity, multi-joint angular configuration, and postural stability. Two modules were developed: a personalized optimal throwing trajectory model that combines historical high-quality samples with the minimum jerk criterion, and a motion deviation diagnosis and recommendation model based on z-scores and hierarchical logic. A total of 2,396 throwing samples from professional and non-professional athletes were collected. Results show that the system generates smooth personalized reference trajectories consistent with natural human movement. Case studies indicate that it can detect poor trunk stability, abnormal elbow displacement, and imbalanced velocity control, then provide targeted recommendations. The framework shifts dart evaluation from deviation from a uniform standard to deviation from an individual's optimal control range, improving personalization and interpretability for darts training and other high-precision target sports.
Figures & tables
Figure 1: Four Factors Affecting Darts Throwing (A) Postural Stability: Demonstrates the supporting role of global postural control. By establishing a stable foundation and minimizing the sway of the body’s center of mass and trunk, it effectively suppresses the error amplification effect transmitted from lower-body disturbances to the upper-body kinetic chain. (B) Joint Angles: This section explains the spatial characteristics of the shoulder joint’s pitch angle, elbow joint’s extension angle, and wrist joint’s flexion angle at the moment of release. The angular configuration of these key joints directly determines the direction of the dart’s initial velocity vector; even the slightest angular disturbance can cause significant deviations in the flight trajectory. (C) Three-Link Motion Model: This model abstracts the throwing arm as a mechanical system composed of the upper arm (Link 1), forearm (Link 2), and hand (Link 3). It intuitively illustrates the physical framework and spatial coordination of the multi-link mechanisms—including the shoulder, elbow, and wrist—throughout the complete throwing cycle. (D) Release Velocity: This reveals the progressive energy transfer pattern based on the kinetic chain. Momentum strictly follows a "proximal-to-distal sequence," starting from the shoulder, transmitting through the elbow joint to the wrist joint, thereby achieving the most efficient power output at the moment of release at the distal end.
Athlete
Professional athletes
Non-professional athletes
S1
S2
S3
S4
S5
S6
S7
Gender
Female
Female
Male
Male
Male
Male
Male
Height
170 cm
168 cm
168 cm
175 cm
176 cm
173 cm
170 cm
Weight
58 kg
57 kg
55 kg
70 kg
55 kg
55 kg
60 kg
Arm Span
165 cm
170 cm
172 cm
180 cm
173 cm
165 cm
168 cm
Num.
237
251
255
305
696
430
222
Table 1: Participant information and data collection summary
Figure 2: Data preprocessing and dart-to-bullseye distance measurement pipeline. Figure (a) displays the raw captured image exhibiting significant perspective distortion and tilt. Figure (b) shows the geometrically restored image following an inverse perspective projection using chessboard calibration parameters. Figure (c) illustrates the bullseye localization process, isolating the red outer ring via HSV color space conversion. Figure (d) presents the isolated dart tip mask generated through frame differencing and morphological operations. Finally, Figure (e) demonstrates the completed recognition result, accurately pinpointing both the bullseye center and the dart tip to compute the exact Euclidean landing distance.
Figure 3: Flow chart of optimal trajectory fitting. This figure illustrates the data-driven pipeline for generating a personalized optimal dart-throwing trajectory. The process initiates by aggregating the athlete’s most recent 200 throwing actions from historical data. Subsequently, a high-quality subset of 30 representative actions is extracted using a rigorous weighting scheme that evaluates both dart-to-bullseye accuracy and kinematic smoothness (jerk rating). Finally, these superior actions are synthesized via hyperparameter optimization and refined using a minimum-jerk kinematic model. This automated filtering and fitting mechanism ultimately yields a smooth, physiologically natural reference trajectory tailored to the individual athlete.
Figure 4: Visualization of the optimal trajectories for seven dart players. The figures in this experiment visualize the fitting results. The top four images show the fitting results for professional athletes, while the bottom three show those for non-professional athletes. You can observe each athlete’s throwing process and hand movements. The green segments represent the motion from the wrist to the fingertips, orange segments represent the motion from the elbow to the wrist, and blue segments represent the motion from the shoulder to the elbow. These visualizations of the optimal fitting trajectories effectively illustrate the athletes’ throwing processes, providing valuable references for darts training.
Characteristics
Professional athletes
Non-professional athletes
S1
S2
S3
S4
S5
S6
S7
Static
Release Velocity(m/s)
5.57 ± 0.58
5.31 ± 0.72
5.15 ± 0.57
5.90 ± 0.81
4.67 ± 0.64
6.80 ± 0.91
7.20 ± 0.75
Release Alignment Angle( ∘ )
13.74 ± 3.40
18.11 ± 3.23
13.54 ± 2.83
13.39 ± 2.62
14.30 ± 2.96
36.28 ± 5.73
13.92 ± 2.42
Mean Grip Distance(mm)
31.26 ± 6.41
75.57 ± 14.73
68.58 ± 12.85
86.09 ± 12.74
52.47 ± 8.49
68.72 ± 11.15
42.50 ± 7.18
Grip Distance Variability(mm)
3.38 ± 0.62
83.19 ± 11.23
36.03 ± 4.61
7.86 ± 1.32
5.53 ± 0.79
70.30 ± 9.26
12.82 ± 1.98
Head Stability Index(mm)
3.45 ± 0.36
2.32 ± 0.20
5.18 ± 0.52
3.11 ± 0.31
5.24 ± 0.52
1.64 ± 0.15
2.06 ± 0.14
Table 2: Table of the 12 Static and 6 Dynamic Characteristics of Darts
Figure 5: Fitted optimal throwing trajectory for athlete S3. This figure shows the time series of the personalized optimal throwing trajectory generated using the minimum jitter fitting method (from frame 15 to frame 120). The red curve depicts a smooth and continuous spatial path of the hand in the XY plane, indicating no excessive trembling or unnecessary fluctuations. Consistent with the principles of the three-link kinematic model, the athlete’s shoulder remains highly stable throughout the entire movement. During the initial backward swing phase (frames 15 to 45), the elbow joint remains in a relatively fixed position to ensure mechanical stability. Subsequently, during the forward acceleration and release phases (frames 60 to 105), the elbow naturally rises to extend the acceleration path, ensuring efficient energy transfer.
Figure 6: Not standard action from S6. This figure presents a non-standard throwing motion captured from a non-professional athlete across six sequential frames. The overlaid skeletal tracking highlights multiple kinematic deviations compared to an ideal reference trajectory. For instance, the sequence visually reveals inconsistencies such as noticeable trunk sway, inadequate head stability, and abnormal elbow displacement during the backward swing and acceleration phases. This specific motion data serves as the direct input for the personalized recommendation model. The system then calculates Z-score deviations across fundamental features to automatically generate targeted corrective feedback for the athlete.
No.
Recommendation
1
Significant trunk instability and standard deviation during movement ( ∣z∣max=11.55 ): Focus on training movement consistency during the 100 ms prior to release.
2
Insufficient head stability during dart release ( z=7.59 ): Fix your gaze on the target, avoid nodding or side-to-side head movement, and maintain the same line of sight before and after release.
3
Insufficient trunk stability during dart release ( z=6.92 ): Engage your core, minimize body sway, and keep your center of gravity stable over your supporting foot.
4
Shoulder Forward Swing Angle ( ∣z∣max=6.40 ): Control arm elevation to align the upper arm’s trajectory with the target; avoid shrugging, which causes path deviation, and use the shoulder as a stable anchor point.
5
Abnormal elbow joint movement in the three-link mechanism ( ∣z∣max=4.66 ): The elbow should remain largely stationary during the backswing; only at a specific point during the forward swing—as the arm accelerates—should the elbow rise naturally, fully extending along a straight line upon release.
6
Acceleration curve is not smooth ( ∣z∣max=4.65 ): Maintain a continuous and even acceleration rhythm; avoid “stopping and then exerting force.”
Table 3: Suggested sentences generated based on the characteristics of the S6 dart
A punch is a ballistic, full-body action driven by a kinetic chain running from the legs through the trunk to the arm, where a small sequencing error separates a scoring strike from a miss. Wearable sensors can capture this movement in the gym, but most deployable systems only classify which punch was thrown rather than assess how well it was thrown. We present Skill Profiling with Attributable Reasoning (SPAR), an eight-IMU garment and pressure-insole system that classifies each punch as expert or novice and treats an explanation of that prediction as feedback. Feedback is only useful if the person receiving it can act on it, so SPAR explains the prediction at three tiers, a per-joint attribution for the analyst, a counterfactual over kinetic-chain layers for the coach, and a plain-language narrative of the two for the athlete. Across 17 participants and 4,713 punches, SPAR reaches a leave-one-participant-out AUC of 0.842 (95% CI [0.769, 0.907] over participants). A frozen time-series foundation model encodes the joint-angle and plantar-force series, and a small transformer trained on the cohort classifies the encoding. We audit the two quantitative tiers and report six themes from a thematic analysis of interviews with six practicing boxing coaches.
Nibraas Khan, Hanchen David Wang, Enya Bullard +5
Department of Computer Science, Vanderbilt University, Nashville, TN 37235, USA · Department of Electrical Engineering, Vanderbilt University, Nashville, TN 37235, USA · Independent Researcher +2
Elite athletics throwing provides a structured biomechanical action sequence together with an objective continuous outcome, namely throw distance, making it a suitable setting for performance-oriented sports analytics. In this work, we investigate whether explicit phase alignment of pose dynamics improves throw-distance prediction under realistic broadcast conditions. To support this study, we curate Throwing4, a dataset of 500 elite throwing clips spanning javelin, discus, hammer, and shot put, with corresponding two-dimensional pose sequences, frame-level phase annotations, and official distance labels. We formulate throw-distance prediction from broadcast-derived pose as a regression problem and compare full-sequence modeling against a phase-conditioned approach that partitions each attempt into four technique stages and fuses phase-specific motion representations. Experiments across multiple graph-based pose encoders show a consistent reduction in prediction error when phase structure is incorporated, indicating that phase-aligned modeling provides a more stable representation than treating the entire motion as an unstructured sequence. These findings suggest that explicit biomechanical phase information is a useful inductive prior for pose-based performance modeling in elite athletics throwing. Beyond score prediction, the dataset and protocol provide a basis for future work on fine-grained technique analysis, interpretable sports analytics, and robust modeling from broadcast video.
Department of Electrical Engineering, Indian Institute of Technology Gandhinagar, Gandhinagar, India. · Department of Electronics and Communication Engineering, Madan Mohan Malaviya University of Technology, Gorakhpur, India. · Department of Mathematical Sciences, IISER Berhampur, Berhampur, India.
We argue that high-quality motion data can steer tracking policies toward better optimization trajectories early in training. In this work, we introduce LIMMT (Less Is More for Motion Tracking). To our knowledge, this is the first data-centric study for physics-based humanoid motion tracking. We go beyond simply removing low-quality and erroneous clips, but define motion data quality through three dimensions: physics feasibility, diversity, and complexity. We show that even training with under 3% of AMASS yields better tracking performance than training with the full dataset. We further conduct data cleaning on the estimated web-sourced mocap data. Extensive experiments and analyses validate the effectiveness of our framework.
Yu Guan, Zekun Qi, Chenghuai Lin +7
1Tsinghua University · 2GalBot · 3Shanghai Jiao Tong University +2