Classification of audio files into separate classes can be incredibly useful for many applications. This project aims to identify important audio features which can be used to distinguish an input audio signal as containing either speech or music. These features are used to train two classification models, and the accuracy and efficiency of these models are evaluated and compared against one another to determine which is more appropriate for different potential applications. Highlighted is the ability of these different training models to handle both higher and lower dimensional datasets.
When downloading the datasets, some data might come in an mp3 format. This needs to be converted .wav using wav_converter.py