Development of a Human Activity Recognition System for Ballet Tasks

Abstract

Background: Accurate and detailed measurement of a dancer's training volume is a key requirement to understanding the relationship between a dancer's pain and training volume. Currently, no system capable of quantifying a dancer's training volume, with respect to specific movement activities, exists. The application of machine learning models to wearable sensor data for human activity recognition in sport has previously been applied to cricket, tennis and rugby. Thus, the purpose of this study was to develop a human activity recognition system using wearable sensor data to accurately identify key ballet movements (jumping and lifting the leg). Our primary objective was to determine if machine learning can accurately identify key ballet movements during dance training. The secondary objective was to determine the influence of the location and number of sensors on accuracy.Results: Convolutional neural networks were applied to develop two models for every combination of six sensors (6, 5, 4, 3, etc.) with and without the inclusion of transition movements. At the first level of classification, including data from all sensors, without transitions, the model performed with 97.8% accuracy. The degree of accuracy reduced at the second (83.0%) and third (75.1%) levels of classification. The degree of accuracy reduced with inclusion of transitions, reduction in the number of sensors and various sensor combinations. Conclusion: The models developed were robust enough to identify jumping and leg lifting tasks in real-world exposures in dancers. The system provides a novel method for measuring dancer training volume through quantification of specific movement tasks. Such a system can be used to further understand the relationship between dancers' pain and training volume and for athlete monitoring systems. Further, this provides a proof of concept which can be easily translated to other lower limb dominant sporting activities Key PointsDeep learning models were shown to have acceptable accuracy when applied to recognised ballet-specific jumping and leg lifting tasks in a population of 23 dancers. A system of multiple sensors (six per dancer) was shown to have the greatest accuracy; however, the optimal single sensor model also performed with acceptable accuracy. The inclusion of all six sensors yielded the highest degree of accuracy: however, fewer sensors still provided an acceptable degree of accuracy. For real-world application, minimal sensors are required to reduce athlete burden. The method demonstrated for model development is highly translatable for future developments in other lower limb dominant sporting activities.

Human Activity Recognition for People with Knee Osteoarthritis—A Proof-of-Concept

“…Usually, the accuracy of HAR algorithms reduces for finer levels of classification as predictions become more complex. Our results are consistent with these previous studies, where reductions in accuracy for subsequent level classification have been reported to range between 4.2% and 16.4% [ 20 , 24 ]. While we compared the results of our models to previous HAR literature, the considerable heterogeneity between these studies reduces the capacity to make meaningful comparisons [ 16 ].…”

“…In addition, deep learning approaches like CNN are able to handle nonlinear interactions between features, something which is limited when using traditional machine learning approaches where features are defined by the researcher. There are many laboratory studies that have reported the accuracy of classifying physical activities from IMU data [ 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 ] using both traditional and deep learning approaches. Ramanujam et al [ 28 ] provide a review of the most up-to-date computational advances in deep learning for HAR which is beyond the scope of this paper.…”

“…There are multiple studies that have developed HAR models to classify daily activities using the lower limbs (e.g., standing, walking, going up stairs, walking down a hill), with accuracy ranging from 83% to 98% using training data from healthy people [ 12 , 13 , 14 , 15 , 16 , 18 , 20 , 25 , 26 , 27 ]. Other studies have classified specific phases of activities like transitioning from sit to stand or the stance phase of walking, reporting accuracy >99% [ 22 , 23 ].…”

Acti-DM1: Monitoring the Activity Level of People With Myotonic Dystrophy Type 1 Through Activity and Exercise Recognition

“…Cheng et al [20] obtained an accuracy of 0.959 in the recognition of specific gym exercises such as bench dips using both accelerometer and gyroscope. More recently, the team of Hendry et al [25] proposed an activity recognition system using 6 sensors to detect simple key movements (jumping and leg lifting) in ballet with the help of a convolutional neural network (CNN) achieving an accuracy of 0.982. Another research team [26] investigated the recognition 10 cross-fit specific exercises based on a deep learning approach (CNN) and reached an impressive recognition rate of more than 99.9 %.…”

Gradient local auto-correlation features for depth human action recognition

“…A large number of researchers have been attracted to human action classification problem due to its wide range of real-world applications. The notable implementations cover visual surveillance [1], smart homes [2], sports [3], entertainment [4], healthcare monitoring [5], patient monitoring [6], elderly care [7], Virtual-Reality [8], human-computer interaction [9], and so on.…”