Are the methods from the urban sounds paper already implemented in Github somewhere? I'm looking for an alternative approach regarding reading audio, stft, pca, whitening and spherical-kmeans since currently my centroids contain lots of noise and I need to check this.