EXPERIMENTAL PHONETICS AND MODELING OF THE SPEECH PERCEPTION SYSTEM

 The goal of experimental-phonetic research is to get an exhaustive indicative description of the sounds of speech, analysing the acoustic signal. In this case, the phonetician knows in advance the “linguistic” structure of the analysed statement and measures the corresponding physical parameters of the signal. In addition, he analyses the speech signal post factum and, if necessary, has the opportunity to repeatedly access the analysis of the same segment.

The purpose of creating a functional model of the speech perception system is to establish a set of features that are necessary and sufficient for an adequate understanding of the utterance, and to create algorithms presumably reproducing the procedures for processing the speech signal by the human auditory system. The speech signal is analysed by the listener on-line and “bottom-up”, i.e. all the information for making linguistic decisions is extracted from the original acoustic signal, and the decisions themselves are made on the basis of the selected parameters and taking into account the knowledge of the language and the current communicative situation available to the listener.

The article gives some examples demonstrating the ability of the perception system when making “linguistic” decisions to dispense with some features from the whole variety of the features of the acoustic realisation of speech sounds gathered today by experimental phonetics.

The methods of experiments used in perceptual phonetics and psycholinguistics do not allow us to use their results in the modelling of speech perception processes.