This is a thesis reproduction project, reproducing the paper: A NOISE-ROBUST FFT-BASED SPECTRUM FOR AUDIO CLASSIFICATION. This project implements the music source separation method from the paper.
The project is based on the knowledge of digital processing of signals and is implemented by Matlab, where the signal processing toolkit of Matlab (NSL Matlab Toolbox) is called.
S1 Calculation of the auditory spectrum
S2 Calculate the auditory spectrogram (power spectrogram) obtained based on the FFT calculation.
S3 Evaluation: Perform feature extraction, which can be divided into three main categories: traditional MFCC feature extraction, MFCC-like features and Spectral features.
S4 Classification with SVM