kyma•tweaky: How SplitConsonantsAndVowels

Question (or Task)

I’d like to split up a speech signal into vowels and consonants. One (pretty crude) way to do it would be to use simply a threshold/gate (the inverse then for the ‘consonants’). But there must be more intelligent ways to accomplish something like this...probably with spectral tracking, as the consonants tend to be more noisy than the vowels... Are there examples for this in the library or on the tweaky? (didn't find anything on first glance) Thanks. Cheers,

-- FlorianWittenburg - 14 Jun 2006

Solution(s)

Here are some examples of separating the voiced from unvoiced parts of live speech:

Kyma Sound Library: An Overview.kym, "Effects-Unvoiced only resynth". Use Threshold to move between voiced/unvoiced.

Kyma Sound Library: Effects>SpectralProcessing-Live> "Lisper" uses a ducker controlled by a consonants-only resynthesis to play a resynthesis with the vowels only, minus the consonants. In that same Sound file, "Unvoiced" plays consonants only.

In Prototypes, take "LiveSpectralAnalysis-Harmonic" and drop "OscillatorBank (replace spectrum)" to its right in the signal flow editor. Uses a checkbox to toggle between voiced unvoiced.

In Tau Editor, Galleries button: TimeFrequency Scaling>"Unvoiced is unscaled...". Set !ScaleFreq to zero for unvoiced; set it to 1 for voiced.

In TauPlayer Sound, check WarpVoicedOnly. Then scale the frequency down to zero.

-- CarlaScaletti - 14 Jun 2006

WebForm
Question:	In a speech signal, how can you separate the voiced from unvoiced parts (or vowels and consonants)
Keywords:	consonants vowels voiced unvoiced speech

k y m a • t w e a k y — the kyma collective || view the current website || February 2015 Archive

Question (or Task)

Solution(s)