Learning Project: Classification and tracking of Sport techniques with Tiny Machine Learning

Karate is a martial art that requires precision, technique, and discipline. To help practitioners improve their skills, a mobile application has been developed that allows users to determine the karate technique performed and provides guidance on its correct execution. This innovative solution combines the power of neural networks and spectral analysis of time series data from a gyroscope. The design and implementation of the classification model is the backbone of this application. By leveraging the capabilities of neural networks, the model can accurately identify different karate techniques based on the data collected from the gyroscope. The gyroscope provides valuable information about the movement patterns and angles, which are crucial for distinguishing between various techniques. To create this model, cutting-edge technologies such as Edge Impulse, Arduino Nano, and App Inventor were utilized. Edge Impulse is a platform that enables the development and deployment of machine learning models on edge devices, making it a perfect fit for this mobile application. The Arduino Nano, a compact and versatile microcontroller, was used to collect the gyroscope data. App Inventor, a visual programming environment, was employed to build the user-friendly interface of the mobile application. The process begins with the collection of training data. Skilled karate practitioners perform each technique while wearing the gyroscope-equipped Arduino Nano. The gyroscope captures the motion data, which is then processed and analyzed using spectral analysis techniques. This analysis helps extract meaningful features from the time series data, which are fed into the neural network for training. The neural network is trained using a large dataset of labeled karate techniques. The model learns to recognize the patterns and characteristics unique to each technique, enabling it to make accurate predictions. The training process involves optimizing the network's parameters to minimize errors and improve its performance. Once the model is trained, it is integrated into the mobile application. Users can record their own karate movements using their smartphones' gyroscopes. The application then processes the recorded data and sends it to the neural network for classification. Within seconds, the user receives feedback on the technique performed and detailed instructions on how to improve their execution. This mobile application has the potential to revolutionize karate training. It provides practitioners with real-time feedback and personalized guidance, allowing them to refine their techniques and enhance their overall performance. By combining the power of neural networks and spectral analysis, this innovative solution opens up new possibilities for improving martial arts training. In conclusion, the development of a mobile application that utilizes neural networks and spectral analysis for karate technique classification is a significant advancement in the field of martial arts training. With this application, practitioners can receive immediate feedback and guidance, leading to continuous improvement and mastery of their skills. The integration of cutting-edge technologies like Edge Impulse, Arduino Nano, and App Inventor makes this solution both accessible and effective.