The Source-Filter Model of Speech
The source-filter model explains the speech signal as the product of a sound source modified by the resonant filtering of the vocal tract.
Definition
The acoustic model of speech production in which a sound source is shaped by the frequency-dependent filtering of the vocal tract to yield the observed speech spectrum.
Scope
This topic presents the acoustic theory in which speech results from two largely independent components: a source—most often the periodic vibration of the vocal folds, or aperiodic turbulence and transient noise—and a filter formed by the resonances of the vocal-tract cavities. The filter's resonant peaks become the formants observed in the spectrum, and changing the vocal-tract shape changes the formant pattern while the source remains the same. The topic explains how this model accounts for vowel and consonant acoustics. The treatment is descriptive.
Core questions
- What constitutes the source in speech, and how does it differ across sound types?
- How does the vocal tract act as an acoustic filter?
- How do formants arise from vocal-tract resonances?
- Why are source and filter treated as largely independent?
Key theories
- Acoustic theory of speech production
- Fant's formalization showing that the output spectrum equals the source spectrum multiplied by the vocal-tract transfer function, so that articulatory changes alter formant frequencies independently of the glottal source.
History
The model was developed in the mid-twentieth century, most influentially in Gunnar Fant's 1960 Acoustic Theory of Speech Production, building on earlier acoustic and electrical-analog studies of the vocal tract. It became the standard framework for relating articulation to acoustics and underlies speech synthesis and analysis methods.
Debates
- Degree of source-filter independence
- Although source and filter are treated as independent, researchers note that in real speech they interact—for instance, vocal-tract configuration can affect glottal vibration—prompting discussion of when the simplifying assumption holds.
Key figures
- Gunnar Fant
- Kenneth Stevens
- Keith Johnson
Related topics
Seminal works
- fant1960
- stevens1998
Frequently asked questions
- What is the 'source' in the source-filter model?
- The source is the sound energy that enters the vocal tract, usually the buzzing of the vibrating vocal folds for voiced sounds, or turbulent noise for fricatives and transient bursts for stops.
- How are formants related to the filter?
- Formants are the resonant peaks of the vocal-tract filter. As the tongue and lips change the shape of the tract, the resonances shift, producing the different formant patterns that distinguish vowels.