The project was done as part of Pattern recognition & Machine learning Master course in UPC Barcelona in year 2017-18.

The project goal was to identify combinations of box punches. The data comes from two smartphones accelerometers. The implementation consists of extracting relevant features, training different classifiers with 6 different classes and testing classifiers on new data. The classes are:

• Jab (Left Straight) - class 1
• Cross (Right Straight) - class 2
• Left Hook - class 3
• Right Hook - class 4
• Left Uppercut - class 5
• Right Uppercut - class 6

Features used for each acceleration axis (6 axes) are:

• statistic - min, max, median, mean and standard deviation
• wavelet - 5 squared detail coefficients of Daubechies 3 wavelet. Libary used for obtaining wavelet features is PyWavelets.

• File cross_validation_comparison.py is used for cross-validation and determining the best classifier and features for our project.
• File testing_set_validation.py is used for testing the new data combinations with best classifier (K-nearest).

• Report of the project is located in the .pdf file