In a sound reproduction or recording system an audio signal is multiplied
by a gain factor (z) which is dependent on the input level (y). The
dependence of the gain factor on input level is chosen such that unvoiced
phonemes are at least 6 dB, preferably at least 12 dB more enhanced than
voiced phonemes, where preferably the average gain is less than 6 dB.
This improves the intelligibility.