A hybrid model for timbre perception : quantitative representations of sound color and density