Speech Production and the Anatomy of Speech

Human speech is produced by coordinating breathing with structures in the throat, mouth and nose. The process begins in the lungs, which push air upward through the windpipe. At the top of the windpipe sits the larynx, or voice box, containing the vocal cords. When air passes between the vocal cords and they are drawn close together, they vibrate and create the basic sound of the voice; tightening or loosening them changes the pitch.

The sound from the larynx is only a buzzing tone until it is shaped by the spaces above, known together as the vocal tract. The throat, the mouth and the nasal passages act as resonating chambers, while the tongue, lips, teeth and the soft and hard parts of the roof of the mouth move to form the distinct sounds of language. Consonants, for example, are made by partly or fully blocking the airflow with the tongue, teeth or lips, while vowels are shaped mainly by the position of the tongue and the openness of the mouth. The sinuses and nasal cavity add resonance and produce nasal sounds.

Speech technology applies this understanding to machines. Speech synthesis generates artificial spoken output, often by modelling how the vocal tract shapes sound, while speech recognition converts spoken words into text or commands. Both fields rely on knowledge of how the anatomy of the mouth and throat creates the patterns of sound that listeners interpret as language.

Frequently asked questions

Where in the body is the voice created?

In the larynx, or voice box, at the top of the windpipe. Air from the lungs makes the vocal cords vibrate, producing the basic sound of the voice.

How are different speech sounds formed?

The tongue, lips, teeth, palate and nasal passages shape the sound from the larynx, blocking or directing airflow to make the consonants and vowels of language.

What is the difference between speech synthesis and speech recognition?

Synthesis produces artificial spoken output from text, while recognition does the reverse, converting spoken words into text or commands.