Archery Score Prediction Project

AIM: Finding the scores of archery shootings using EEG Signals

🎯 Archery EEG Signals Analysis

EEG Signal Analysis for Archery Performance Prediction

📋 Project Overview

This research focuses on predicting archery shooting scores using EEG (Electroencephalogram) signals. The project analyzes brain activity data from archers during their shots to predict performance metrics. By leveraging machine learning and signal processing techniques, we aim to understand the relationship between neural activity and archery performance.

Research Context

• EEG signals provide non-invasive insights into brain function during archery shots
• Analysis of different frequency bands (Delta, Theta, Alpha, Beta) reveals distinct mental states
• Focus on correlating specific brain activities with archery performance

📊 Dataset

The dataset, provided by Eskişehir Technical University Sports Science Faculty, consists of:

1. EEG Signals with Scores:

   • EEG recordings for the last second of each shot (250 Hz frequency)
   • Corresponding performance scores

2. Score vs. Distance Data:

   • Shot scores
   • Arrow distances from target center

🔬 Methodology

Data Preprocessing

• Merging EEG signals with score/distance data
• Distance correction and normalization
• Feature extraction from EEG frequency bands:
  • Delta (4-8 Hz)
  • Theta (8-12 Hz)
  • Alpha (12-20 Hz)
  • Beta (20-30 Hz)

Models Implemented

1. Random Forest Regressor

   • Robust and interpretable
   • Feature importance analysis
   • Hyperparameter tuning via GridSearchCV

2. LSTM (Long Short-Term Memory)

   • Specialized for sequential data
   • Optimized through RandomizedSearchCV
   • Effective for time-series EEG analysis

3. Support Vector Regressor (SVR)

   • Non-linear relationship handling
   • Radial basis function kernel
   • Hyperparameter optimization

📈 Results

Model Performance Metrics

• Mean Squared Error (MSE)
• R-squared (R2)
• Mean Absolute Percentage Error (MAPE)

Key Findings

• Random Forest achieved highest R² score (0.083226) after tuning
• LSTM showed lowest MAPE (0.236162) among all models
• Beta frequency band (12-20 Hz) identified as most significant for prediction

🛠️ Tools and Technologies Used

Data Processing & Analysis

• Python - Primary programming language
• NumPy - Numerical computing and array operations
• Pandas - Data manipulation and analysis
• SciPy - Scientific computing and signal processing

Machine Learning & Deep Learning

• TensorFlow - Deep learning framework for LSTM models
• Scikit-learn - Machine learning library for:
  • Random Forest Regression
  • Support Vector Regression (SVR)
  • Model evaluation metrics
  • Data preprocessing
• StandardScaler - Feature scaling and normalization
• GridSearchCV - Hyperparameter tuning

Signal Processing

• SciPy FFT - Fast Fourier Transform for frequency analysis
• SciPy Welch - Power spectral density estimation
• Signal filtering and transformation tools

Data Visualization

• Matplotlib - Data visualization and plotting
• Seaborn - Statistical data visualization
• Interactive plotting tools

Development Tools

• Jupyter Notebooks - Interactive development and analysis
• Git - Version control
• Virtual Environment - Dependency management

🚀 Future Work

• Collect more extensive and balanced dataset
• Include data from wider range of archers and conditions
• Enhance model robustness and generalizability
• Implement advanced feature extraction techniques

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

📚 References

1. Vendrame, E., et al. (2022). Performance assessment in archery: a systematic review. Sports Biomechanics, 1-23.
2. Khosla, A., et al. (2020). A comparative analysis of signal processing and classification methods for different applications based on EEG signals. Biocybernetics and Biomedical Engineering, 40(2), 649-690.