A Robust Deep Learning Framework for Skill Level Discrimination in Tennis Strokes Using Bilateral IMU Measurements.

Area of Science:

• Sports Science
• Biomechanics
• Machine Learning

Background:

• Objective skill classification is crucial in tennis for coaching and talent identification.
• Performance relies on complex kinetic chain interactions, making visual assessment subjective.

Purpose of the Study:

• To develop a deep learning framework for objective tennis skill classification using Inertial Measurement Unit (IMU) data.
• To distinguish between elite and amateur tennis players based on kinematic stroke analysis.

Main Methods:

• A hierarchical deep learning framework utilizing synchronized bilateral IMU data from 39 participants.
• A hybrid Convolutional Neural Network-Bidirectional Long Short-Term Memory (CNN-BiLSTM) architecture for spatiotemporal feature extraction.
• Analysis of forehand, backhand, service, and volley strokes from 4594 samples.

Main Results:

• The CNN-BiLSTM model achieved 95.54% accuracy in distinguishing expertise levels.
• The system outperformed traditional machine learning and existing deep learning benchmarks.
• Elite athletes showed homogeneous clusters in feature space, and professionals demonstrated superior bilateral coordination stability.

Conclusions:

• The proposed deep learning system provides a robust, field-applicable solution for identifying technical excellence in tennis.
• Digital biomarkers derived from IMU data overcome the limitations of subjective visual observation in coaching.
• This technology offers reliable tools for coaches in talent identification and performance analysis.