Automatic recognition and scoring of olympic rhythmic gymnastic movements

Evaluation of two different 1-year training programs among prepuberal female children

The aim of this study was to analyse and compare the effects of two different dance training concepts in terms of anthropometric indices and supervised motor test performance between two groups of 56 prepuberal girls. The first group, composed of 29 girls, performed a typical classical dance training program (D group), while the second one, composed of 27 girls, followed a mixed training (D&R group) workout, with classical dance exercises alternated to rhythmic gymnastics. Anthropometric and exercise performance data were collected and assessed initially (T1) and after a 1-year (T2) training program. Data collected were firstly tested for collinearity using Belsley’s test and, after their normality was tested through the Shapiro–Wilk test, analysis of variance (ANOVA) tests were performed on a subset of noncollinear variables. The two groups showed improvements in both anthropometric parameters and exercise performance; however, no differences arose between the variation of anthropometric parameters between the two groups. On the other hand, exercise performance was improved in some fields of the D&R group, particularly in exercises regarding the upper body, such as pull-ups, chin-ups and sit-ups. In conclusion, our data provide evidence that a specific training program consisting of classic dance plus rhythmic gymnastics exercises (D&R) has a greater impact compared to a program of simple classic dance, in terms of skill performance without impairment of anthropometric parameters.

Longitudinal Study of Individual Exercises in Elite Rhythmic Gymnastics

The performance evolution in rhythmic gymnastics depends on changes in code of points. At the beginning of each Olympic cycle the code of points changes and therefore, the content of the competition exercises, as well. This study aimed to analyze - for each apparatus - the evolution of number of technical elements and final score over the last two decades (last 13 world championships), how they have been affected by changed code of points, and how the final score relates to the number of technical elements performed. The sample consisted of 416 exercises in five apparatus: ball (96), rope (40), hoop (96), ribbon (88), and clubs (96). The following variables were gathered: code of points, apparatus, technical group, total number of elements, final classification, and final score. Linear mixed-effects models were used to examine the effects on the number of elements and final score in each apparatus. The number of technical elements increased in all apparatus, between 7.4 and 20% over a 10-year period. There were mixed evolutions of final score between the different apparatus, between -6.3 and 14% over a 10-year period. There is small increase in number of elements in hoop and a small decrease in rope after a code change. There was a small decrease in final score in championships after a code change in hoop, moderate in clubs and ribbon, and large in rope. There was a negative relationship between number of elements performed and final score in clubs. In conclusion, the code change generally effects the final score negatively, but there were apparatus specific effects of code change on number of elements and relationship between number of elements and final score.

Analysis of Movement Quality in Full-Body Physical Activities

Full-body human movement is characterized by fine-grain expressive qualities that humans are easily capable of exhibiting and recognizing in others’ movement. In sports (e.g., martial arts) and performing arts (e.g., dance), the same sequence of movements can be performed in a wide range of ways characterized by different qualities, often in terms of subtle (spatial and temporal) perturbations of the movement. Even a non-expert observer can distinguish between a top-level and average performance by a dancer or martial artist. The difference is not in the performed movements--the same in both cases--but in the “quality” of their performance.

In this article, we present a computational framework aimed at an automated approximate measure of movement quality in full-body physical activities. Starting from motion capture data, the framework computes low-level (e.g., a limb velocity) and high-level (e.g., synchronization between different limbs) movement features. Then, this vector of features is integrated to compute a value aimed at providing a quantitative assessment of movement quality approximating the evaluation that an external expert observer would give of the same sequence of movements.

Next, a system representing a concrete implementation of the framework is proposed. Karate is adopted as a testbed. We selected two different katas (i.e., detailed choreographies of movements in karate) characterized by different overall attitudes and expressions (aggressiveness, meditation), and we asked seven athletes, having various levels of experience and age, to perform them. Motion capture data were collected from the performances and were analyzed with the system. The results of the automated analysis were compared with the scores given by 14 karate experts who rated the same performances. Results show that the movement-quality scores computed by the system and the ratings given by the human observers are highly correlated (Pearson’s correlations r = 0.84, p = 0.001 and r = 0.75, p = 0.005).

Machine and deep learning for sport-specific movement recognition: a systematic review of model development and performance

Objective assessment of an athlete's performance is of importance in elite sports to facilitate detailed analysis. The implementation of automated detection and recognition of sport-specific movements overcomes the limitations associated with manual performance analysis methods. The object of this study was to systematically review the literature on machine and deep learning for sport-specific movement recognition using inertial measurement unit (IMU) and, or computer vision data inputs. A search of multiple databases was undertaken. Included studies must have investigated a sport-specific movement and analysed via machine or deep learning methods for model development. A total of 52 studies met the inclusion and exclusion criteria. Data pre-processing, processing, model development and evaluation methods varied across the studies. Model development for movement recognition were predominantly undertaken using supervised classification approaches. A kernel form of the Support Vector Machine algorithm was used in 53% of IMU and 50% of vision-based studies. Twelve studies used a deep learning method as a form of Convolutional Neural Network algorithm and one study also adopted a Long Short Term Memory architecture in their model. The adaptation of experimental set-up, data pre-processing, and model development methods are best considered in relation to the characteristics of the targeted sports movement(s).

Biomechanical metrics of aesthetic perception in dance

The brain may be tuned to evaluate aesthetic perception through perceptual chunking when we observe the grace of the dancer. We modelled biomechanical metrics to explain biological determinants of aesthetic perception in dance. Eighteen expert (EXP) and intermediate (INT) dancers performed développé arabesque in three conditions: (1) slow tempo, (2) slow tempo with relevé, and (3) fast tempo. To compare biomechanical metrics of kinematic data, we calculated intra-excursion variability, principal component analysis (PCA), and dimensionless jerk for the gesture limb. Observers, all trained dancers, viewed motion capture stick figures of the trials and ranked each for aesthetic (1) proficiency and (2) movement smoothness. Statistical analyses included group by condition repeated-measures ANOVA for metric data; Mann-Whitney U rank and Friedman's rank tests for nonparametric rank data; Spearman's rho correlations to compare aesthetic rankings and metrics; and linear regression to examine which metric best quantified observers' aesthetic rankings, p < 0.05. The goodness of fit of the proposed models was determined using Akaike information criteria. Aesthetic proficiency and smoothness rankings of the dance movements revealed differences between groups and condition, p < 0.0001. EXP dancers were rated more aesthetically proficient than INT dancers. The slow and fast conditions were judged more aesthetically proficient than slow with relevé (p < 0.0001). Of the metrics, PCA best captured the differences due to group and condition. PCA also provided the most parsimonious model to explain aesthetic proficiency and smoothness rankings. By permitting organization of large data sets into simpler groupings, PCA may mirror the phenomenon of chunking in which the brain combines sensory motor elements into integrated units of behaviour. In this representation, the chunk of information which is remembered, and to which the observer reacts, is the elemental mode shape of the motion rather than physical displacements. This suggests that reduction in redundant information to a simplistic dimensionality is related to the experienced observer's aesthetic perception.