Quantitative analysis of the effects of emphasis upon prosodic features of speech

Abstract

Abstract While it is known that emphasis is represented mainly by fun-damental frequency, speech rate, and source intensity, few stud-ies have been published on the relative roles of these variablesin expressing the degree of emphasis. The present paper intro-duces the relative speech rate and the relative source intensity ofa target utterance against a reference utterance, and formulatesthe processes of their generation by quantitative models that arein line with the model that has been established for the funda-mental frequency contour. This makes it possible to comparethe effects of emphasis on the three variables in quantitativeterms, as well as to compare the effects of various degrees ofemphasis. Analysis of English utterances by a native speakerand a non-native speaker indicated the influence of emphasis onthe three variables in quantitative terms, and also clarified thedifference between native and non-native speakers. 1. Introduction It is well known that various prosodic features contribute tothe expression of both linguistic and paralinguistic information.Here we define linguistic information as the symbolic informa-tion that is represented by a set of discrete symbols and a setof rules for their combination. We also define paralinguistic in-formation as the information that is not implicit in the linguis-tic symbols, but is deliberately added by the speaker to modifyor supplement linguistic information [1]. Paralinguistic infor-mation thus defined generally has both discrete and continuousaspects. For example, the information regarding whether a spe-cific part, say a syllable or a word, is emphasized or not is dis-crete, but the information regarding the degree of emphasis iscontinuous. In other words, the speaker can express both thecategory ‘emphasis’ and the degree within the category. Thepresent study takes up the emphasis as an example of paralin-guistic information.In many languages it is generally accepted that pitch, quan-tity, and loudness are the three major subjective features re-lated to the expression of emphasis, corresponding to objectiveprosodic variables of fundamental frequency, speech rate, andintensity, respectively. It is also known that languages may dif-fer in the relative importance they put on these three variablesin expressing emphasis. Comparatively little has been known,however, on the quantitative aspects of emphasis, i.e.,onhowthe degree of emphasis is expressed — whether or not the threeprosodic variables contribute equally to expressing the degreeof emphasis. In order to obtain clear understanding on this is-sue, we need to have methods for quantitative analysis of thesethree prosodic variables.The present paper first describes the methods we developedto obtain quantitative measures for the effects of the degree ofemphasis intended by the speaker on the three prosodic vari-ables, and then presents some results of comparison betweenutterances ofa native speakerand anon-native speakerofAmer-ican English in expressing the degree of emphasis.