Kozminski Techblog

Reading Time: 4 minutes

As we all know, analyzing a player’s technique in professional sports is very important and common. My post was based on quite old articles related to AI in sports. To this day, Kohonen’s maps are not used very often, but this blog aims to show us how in 1995 people were aware that artificial intelligence would have an impact even in this field. However, it is worth stopping at this post and realizing how, despite a few accurate observations, technology surprises us and overtakes our speculations, and the question arises: Will what we imagine now really look like this in our future?

Introduction

Over the last decade, the integration of Artificial Intelligence (AI) in sports biomechanics has witnessed significant strides, with advancements in Expert Systems, Artificial Neural Networks (ANNs), and Evolutionary Computation. In that post I will reflect on the developments in the field, comparing the use of Expert Systems in gait analysis with their limited presence in sports biomechanics. I will also delve into the applications of ANNs, specifically Kohonen self-organizing maps, and explores the emerging role of Evolutionary Computation in optimizing sports techniques.

Expert Systems: A Slow Start

In 1995, Lapham and Bartlett predicted a promising future for Expert Systems in sports biomechanics. However, a decade later, these systems, essentially a combination of a database, knowledge base, reasoning, and a user interface, are still underutilized. Unlike in gait analysis, where Expert Systems are employed for diagnostic purposes, their implementation in sports biomechanics has been scarce. The reluctance may stem from the complexity of technique analysis, the lack of a strong developmental motivation, and the challenges of dealing with fuzzy, imprecise data.

Expert Systems in Sports Biomechanics

Expert Systems, powerful knowledge databases, hold immense potential in transforming sports biomechanics. In a cricket context, a hypothetical expert system for fast bowling might use rules like: IF “shoulder-axis counter-rotation” is high; THEN “technique” is mixed (p = 0.8). Handling the vagueness in biomechanical data, exemplified in Figure 1 for fast bowling, showcases the challenge. These systems act as robust diagnostic tools, offering valuable insights and aiding in error identification for flexible sports techniques.

Utilizing expert systems, possibly integrated with video analysis tools like SiliconCOACH’s ‘wizards,’ holds promise for developing diagnostic tools to identify technique errors. This aligns with the optimistic view on the utility of expert systems in sports biomechanics expressed by Lapham and Bartlett in 1995.

Artificial Neural Networks (ANNs): Mapping Movement Patterns

In contrast to Expert Systems, ANNs, especially Kohonen self-organizing maps, have found a niche in sports biomechanics. ANNs mimic the brain’s neural network, allowing computers to learn from experience and analyze complex movement patterns. Studies utilizing Kohonen maps have shown promise in discerning patterns in discus throws, javelin throws, soccer kicks, and more. Despite their successes, challenges remain in deciphering the output map nodes and determining their relevance to movement characteristics.

Evolutionary Computation: Predicting Optimal Techniques

Evolutionary Computation, incorporating genetic algorithms and evolutionary strategies, has made a notable appearance in optimizing sports techniques. In a soccer throw-in scenario, an evolutionary strategy successfully predicted an optimal technique aligning with coaching literature. This application showcases the potential of Evolutionary Computation in refining sports skills.

Future Perspectives

As technology evolves, automatic marker-tracking systems enable the collection of vast and precise human movement data. This may pave the way for the development of fuzzy Expert Systems for diagnosing faults in sports techniques. Kohonen mapping is expected to become commonplace, provided researchers can identify the specific technique elements captured by these maps. Multi-layer ANNs are anticipated to play a crucial role in technique analysis, building on their success in biomechanics and gait analysis. Evolutionary Computation and hybrid systems are likely to feature prominently in optimizing sports techniques and skill learning.

Conclusion

While recent years have seen remarkable progress in integrating artificial intelligence into sports biomechanics, challenges and untapped potential remain. The optimism expressed in 1995 by Lapham and Bartlett has not yet been fully realized, but with continued progress and a growing understanding of the applications of artificial intelligence in traffic analysis, the future appears promising. The synergy between artificial intelligence and sports biomechanics can open new dimensions in optimizing sports performance, improving sports techniques and avoiding injuries. It is worth noting, however, that considering the great development in the world when it comes to artificial intelligence, sports are still a niche direction in which technology is developing.

LET’S REMEMBER THAT THERE ARE MANY UNDERESTIMATED AREAS THAT ARE JUST WAITING TO DEVELOP- one of them is sport

Resources:

• Artificial Intelligence in Sports Biomechanics: New Dawn or False Hope? – PMC (nih.gov)
• The science and medicine of cricket: an overview and update: Journal of Sports Sciences: Vol 21, No 9 (tandfonline.com)
• The use of artificial intelligence in the analysis of sports performance: A review of applications in human gait analysis and future directions for sports biomechanics: Journal of Sports Sciences: Vol 13, No 3 (tandfonline.com)
• AI in Biomechanics: The Key to Unlocking Human Movement Potential (ts2.pl)
• Self-Organizing Maps for the Analysis of Complex Movement Patterns | Neural Processing Letters (springer.com)

AI used: ChatSonic