Kata, a fundamental component of karate, embodies a series of offensive and defensive movements performed in a structured pattern, showcasing artistic and technical prowess. Evaluating kata performance traditionally faces challenges like human bias and inaccuracies, which emphasize the need for objective evaluation methods. This study aims to develop an artificial intelligence (AI) system to analyze and assess the technical execution of kata movements in karate using image processing and machine learning techniques. The proposed AI system captures high-quality video recordings of athletes performing specific kata, such as Heian Jodan, and employs convolutional neural networks (CNNs) to extract critical motor parameters, including timing, balance, accuracy, and joint angles. It then evaluates performance against predefined technical standards. The system also provides immediate, detailed feedback to players and coaches, highlighting areas of strength and areas requiring improvement. The study was conducted on a sample of 50 brown-belt karate athletes, divided equally into experimental and control groups. The experimental group utilized the AI system for performance evaluation, while traditional coach evaluations assessed the control group. Statistical analyses, including T-tests, revealed that the AI system delivered accurate feedback, closely aligning with the traditional coach assessments, with a minimal difference of 1.11% to 1.19%. Key findings highlighted high accuracy levels (89%-95%) in simpler movements like "Yui (Preparation)" but challenges in maintaining balance during backward movements and slower reaction times in defensive moves. Recommendations include specialized exercises to improve balance, reaction speed, and energy management, alongside integrating the AI system into training programs for consistent and objective performance analysis. This research signifies a technological shift in sports training, offering a reliable, advanced tool to enhance performance evaluation and skill development in karate, bridging gaps in traditional assessment methods while promoting, professional growth in martial arts.
© Copyright 2025 Scientific Journal of Sport and Performance. Asociación Española de Análisis del Rendimiento Deportivo. Alle Rechte vorbehalten.