Audiobia

I chose the audio classification task as the open-topic course project for Machine Learning (CS420). I implemented an audio classification model that adopted pre-trained EfficientNet. To enhance its discernment of audio traits, I extracted harmonic and percussive features from audio data using Harmonic Percussive Source Separation (HPSS), and fed them into EfficientNet for fine-tuning.

• You can see my mini paper here.

Here are some audio spectrum pictures.