Describir: Comparing audio descriptors for singing voice detection in music audio files