Pitch contour analysis

A major impediment to pitch contour analysis is that pitch contour shape has traditionally been analysed qualitatively using human raters (e.g. Femald and Simon, 1984). There have recently been several attempts to find reliable quantitative methods, but no single approach has emerged as either simple in its execution or capable of being applied to a wide variety of speech types. For instance, Trainor et al. (2000) used measurements of pitch values at prespecified points within a contrived stock phrase. This approach was unable to provide a sensitive characterization of pitch contours. By contrast, Moore et al. (1994) successfully applied mathematical modelling to the problem. Then-approach would potentially allow a wide variety of speech types to be addressed, but the complexity of their approach seems to have prevented its wide acceptance within psychology. 

Pitch contour extraction was achieved using the freely available speech analysis program, Praat 4.1.19 (Boersma and Wernicke, 2004). The pitch range for the extraction was set at a floor of 100 Hz and a ceiling of 600 Hz (recommended setting in Praat 4.1.19 for fanale voices). It should be noted that although the whole word was used for the extraction process, only visible pitch contours were extracted (which often do not encompass whole words). The smooth function was then used to smooth the contours (at lO-Hz bandwidth). This is a standard procedure to obtain... 

---