DEFENDED IN 1996 Pitch perception in speech : a time domain approach

This thesis deals with the perception of pitch of speech signals and ways to measure this pitch. The term 'pitch' is used as a collective term for speech production (where it is used for the fundamental frequency of quasi-periodically vibrating vocal folds), for acoustic transmission (where it is used for the periodicity of signals), and for speech perception (where it is used for the perceived pitch of speech signals). In the last case, the frequency of the perceived pitch of a signal is expressed by the fundain�ntal frequency of a reference signal with a rather simple structure (e.g. a pure tone). ' ciften, the values for 'pitch' are identical in speech production, transmission, and perception. But the frequency of the quasi-periodically vibrating vocal folds might be attenuated by the vocal tract shape, and a higher harmonic of the FO can be amplified, which appears as clear periodicity in the acoustic signal and is perceived as pitch. Consequently, pitch and fundamental frequency are not always the same. As argued in Chapter 2, '.Yhat is important for speech communication is what is perceived as pitch and not what is produced as fundamental frequency. Voice source and vocal tract filter information are not linear independent and they are perceived as one signal, rather than being decomposed into separate components. This does not lessen the impact of Fant's (1960) source filter model as an adequate way to describe speech production; but it states that perception is not simply an in version of the production process. From this it follows directly that the motor theory of speech perception (Liberman & Mattingly, 1985) is untenable, because it predicts the perception of the intended gestures that generated a speech sound. The view that source and filter information are perceived together as one entity is compatible with the invariance theory of speech perception (Blumstein & Stevens, 1981), because here all information in a speech signal is used as an integrated percept, without considering how a sound might have been produced. Finally, the proposed view of pitch perception is also in accordance with the auditory theory of speech perception (Seneff, 1985), because here the acoustic signal is perceived as one percept which is represented in two different ways simultaneously. This thesis argues for an integrated