In the [[source-filter model]] of speech production, the source generator is the component of the speech production system that creates the raw acoustic energy before it is shaped into specific speech sounds. In this model, the [[z-transform]] of the speech signal $S(z)$ can be expressed as the product of the source $E(z)$ and the filter $H(z)$: $ S(z) = E(z) H(z) $ The source $E(z)$ provides the excitation to the filter $H(z)$ that models the [[vocal tract]]. Since the human anatomy can produce two primary types of acoustic sources, depending on whether you are producing a vowel, a "hissing" sound like /s/, the source generator must be able to generate both periodic and aperiodic source signals. The periodic source models the quasi-periodic air pulses resulting from the [[vocal folds]] vibration, and is usually called the [[voiced source generator]]. The aperiodic source models air turbulence when it is forced through a narrow constriction in the vocal tract, or sudden releases of pressure after a total closure, and is usually called the [[unvoiced source generator]].