Create synthetic fricatives script (readme file) by Matthew B. Winn February 2012 this script takes a 250 ms file of white noise (at 70 dB) and creates a fricative according to the dimensions that you specify For each of three spectral peaks in the fricative noise, you specify: 1- the center frequency 2- the bandwidth 3- the amplitude relative to the central peak You also set the duration, rise time and fall time of the fricative. You might notice in the script that there are commented-out sections that refer to naming and saving the file. Currently, the script doesn't do this too well - you're welcome to try to fix it up! As it stands, the script should be run to create one fricative at a time, then you can re-name it and save it as you wish. The script is currently set up to run on a PC. Here are the settings that I have found to work for each of four voiceless fricatives in English: fa tha sa sha spp1 1700 2000 4700 2500 spp2 4700 5700 6200 3400 spp3 7700 7600 8200 5200 bw1 3200 3200 1500 2000 bw2 4500 4500 2000 2600 bw3 4000 4000 2500 3500 rel1 5 5 3 8 rel2 0 0 0 0 rel3 -5 -3 -6 6 int 60 55 63 67 the "peaked" option is if you want the center frequencies to peak out a little bit more from the background spectrum. I recommend it. Note: the spectral properties of these fricatives is dependent on: 1- the vowel environment 2- the talker's gender 3- other idiosyncracies of the talker For best results, look at the spectrum of a fricative that you want to model, then use landmarks on that spectrum to set the parameters of your synthesis.