“…Then, to reduce the dimension of our SPL vector, we only keep the SPL( f c ) measures per octave that maximize the Pearson correlation coefficient with the variable W 44013 . The acoustic descriptors resulting from this process, also listed in Table I, form a 10-D feature vector over all the observations consisting of the following descriptors: SPL(0.1), SPL(0.7), SPL(1.8), SPL(5), SPL(8), SPL(11), SPL( 14), SPL (19), SS(2-8), and SS (8)(9)(10)(11)(12)(13)(14)(15). In the following, all f c values are rounded to a multiple of 100 Hz for clarity in the notation, and all SPLs are given in dB ref μPa 2 .…”