The goal of this CMLS homework is to build a classifier that could predict the audio event recorded in an audio excerpt.
We evaluated the audios using these features:
- MFCC mean and standard deviation
- Chroma features mean and standard deviation
- Spectral contrast mean and standard deviation
As suggested by the author of the dataset, we performed the 10-fold cross validation to do training, test and classification steps. Regarding the classification, we decided to use SVM (Support Vector Machine) as model for the classifier and RBF (radial basis function) as kernel function.
[UrbanSound8k] (https://www.kaggle.com/chrisfilo/urbansound8k)
• Giovanni Zanocco • Elisa Castelli • Antonio Rizzitiello • Emanuele Intagliata