TABLE OF CONTENTS Chapter written by Fr d ric BIMBOT.
A spoken message does not merely convey the meaning of the message produced by the person who utters it. It also conveys information about the speaker, and more particularly, his or her identity. Each of us is indeed able to recognize familiar people by their voice.
In general, the main objective of spoken communication is centered on the spoken message. Variability due to the speaker (i.e. the transmitter of the message), is compensated for by the interlocutor (i.e. the receiver).
In speech recognition, a key goal is to neutralize inter-speaker variability. Conversely, in speaker recognition, the focus is put on the characteristics of the transmitter. Inter-speaker variability factors become essential to characterize the voice of a particular speaker. At the same time, intra-speaker variability becomes predominant, and acoustic factors such as noise, microphone or channel variations play a limiting role.
Voice variability between people originates primarily from the existence of morphological differences in speech production organs, in particular the length of the vocal folds, the shape of the oral and nasal cavities and of the various articulators. These elements evolve during childhood, and they stabilize in adulthood, during which they hardly change, even though specific pathologies (such as colds) can modify their characteristics more or less permanently.
Another dimension of inter-speaker variability comes from physiological differences which impact motor control, such as muscle tonicity, motor accuracy, movement coordination and,...
