Hillenbrand: Source Filter Theory Source-filter theory can be used to describe and explain any kind of speech – vowels, consonants, periodic, aperiodic, mixed, static, dynamic, you name it. Our initial goal is to understand the application of SFT to just one simple thing: production of a sustained, phonated vowel. We’ll then see how SFT can be used to described anything and everything related to speech. 12/6/2018 Hillenbrand: Source Filter Theory
The Source of Source-Filter Theory (during phonated speech)
Hillenbrand: Source Filter Theory 12/6/2018 Hillenbrand: Source Filter Theory
Hillenbrand: Source Filter Theory 12/6/2018 Hillenbrand: Source Filter Theory
[ɑ] [i] [u]
Source-Filter Independence 12/6/2018 Hillenbrand: Source Filter Theory
Brad Story (University of Arizona) Source-Filter Demos https://www.youtube.com/watch?v=V_EUMiSOdOI https://www.youtube.com/watch?v=LKXmVD1L9Mw 12/6/2018 Hillenbrand: Source Filter Theory
Source-Filter Theory Demonstration: Peter Frampton’s Talk Box http://www.youtube.com/watch?v=g9D-kUEp03c
Questions In what way does the Talk Box illustrate the principle of source-filter independence? In what way does whispered speech illustrate the principle of source-filter independence? Draw a source-filter theory model (source spectrum, VT FRC, and output spectrum) for the production of whispered speech, assuming that the source spectrum for whisper is white noise (which it isn’t, quite, but close enough).
Hillenbrand: Source Filter Theory The source does not have to be the larynx and it does not have to be periodic. For example, the source for unvoiced fricatives is produced in the vocal tract rather than the larynx and is (roughly) white noise. The pictures below show SF models for /s/ and /ʃ/. 12/6/2018 Hillenbrand: Source Filter Theory
/s/ and /ʃ/ SF Models Overlaid 12/6/2018 Hillenbrand: Source Filter Theory
Hillenbrand: Source Filter Theory Previous examples: Unvarying source, unvarying filter. Everyday speech: Dynamic; i.e., source changes in pitch (f0), loudness (intensity), and quality (FRC) over time; filter varies over time as the articulators move. Source-filter theory can accommodate the dynamic nature of speech simply by allowing the pitch, loudness, and detailed shape of the source to vary over time and by allowing the vocal tract FRC (i.e., the VT filter) to vary over time as the articulators move from one position to the next. 12/6/2018 Hillenbrand: Source Filter Theory
Time-Varying Source Spectrum This is what the 1st part of time-varying source for “Where were you…” looks like in the frequency domain. 12/6/2018 Hillenbrand: Source Filter Theory
Time-Varying Source Waveform (i.e., time domain) ... 12/6/2018 Hillenbrand: Source Filter Theory
Time-Varying Vocal Tract Filter Source Only Source Modified by Time-Varying Filter 12/6/2018 Hillenbrand: Source Filter Theory
Sinewave Speech Demonstration Sinewave speech examples (from HINT sentence intelligibility test): 12/6/2018 Hillenbrand: Source Filter Theory
Sinewave Speech Example 12/6/2018 Hillenbrand: Source Filter Theory