Audio: formant (0:05)

Figure: formant

formant plays a sequence of five vowels: a, e, i, o and u. The sound is synthetic speech generated by a computer using a technique known as formant synthesis, one of a number of techniques used for speech synthesis. Each vowel contains 5 formants. Each formant is created by band-pass filtering white noise and the five formants are mixed together to produce the final vowel sound. The formant figure shows the spectrogram of the first vowel, a. You should, just, be able to see five peaks in the spectrum representing the five formants that comprise the vowel sound. The formant centred on 600Hz is the loudest, the next is centred on 1040Hz, then there is a gap to the third and fourth at 2250Hz and 2450Hz, and the quietest is at 2750Hz. The formants do not form a harmonic series.

A formant is a peak in the sound spectrum, a range of frequencies which are louder than their neighbours. Singing produces formants as do musical instruments.

Resonance is the phenomenon that makes formants sound louder than other sounds. Resonance in singing is achieved by the mouth and throat which filter a sound in different ways to make certain bands of frequencies sound louder than others. The range of frequencies and their amplitude depend on the physical characteristics of the singer and the physical size and shape of the vocal tract.

Vowels and consonants both contain formants. The frequency range and amplitude of these formants enables one vowel or consonant to be distinguished from another. A vowel is created with an open vocal tract, a consonant with a constricted or closed vocal tract. A vowel formant sounds louder than a consonant formant because the vocal tract is open. This is the reason singers are often said to sing vowels, as vowel formants sound the loudest.

Every vowel has an associated set of seven, eight or more formants. An interesting fact of acoustics is that the frequency range of the first three formants is all that is needed to enable people to distinguish one vowel from another.

Formant synthesis is one method for producing an artificial voice. Another example is a vocoder. Synthetic speech has been around for a long time but synthetic singing is still in its infancy. Most attempts to date to synthesise singing make the singer sound like a robot but this will undoubtedly improve in the future.