Unsupervised, Auto K-Means Audio Clustering using Dynamic Weight Selection