Wish the following paper can help you a little bit:http://archive.
signalprocessingsociety.org/ technical-committees/list/sl- tc/spl-nl/2013-11/ SpeakerIdentification/Gang Liu--On Mon, Feb 6, 2017 at 12:56 PM, Huron, David <huron.1@xxxxxxx> wrote:Dear collective wisdom,
I recall reading somewhere (long ago) that most human screams exhibit a spectral peak in the 3 kHz region (more broadly, 2-5 kHz) coinciding with the threshold dip due to the ear canal resonance. The implication is that screams are co-adapted to the most sensitive region of human hearing.
Can anyone point me to a suitable reference? Even literature reporting power spectrum data for human screams would be useful.