Intensity in relation to prosody organization

Mandarin fluent speech prosody is most significantly characterized by phrase grouping. A hierarchical prosody framework of phrase grouping has been proposed, where corresponding evidence of governing effects from the prosody organization were found in two acoustic correlates, namely, overall F₀ contours and temporal allocations (Tseng, C. et al., "From Traditional Phonology to Mandarin Speech Processing", Beijing Foreign Language Teaching and Research Process, p.417-38, 2004). We present results of investigating, through corpus analyses, a third acoustic correlate, i.e., intensity, to look for corresponding evidence in relation to prosody organization as well. The specific questions raised are: (1) how the intensity pattern can be explained by the governing effect of prosody organization; (2) whether the governing effect can be used in predicting intensity distribution. We argue that the acoustic roles of speech rhythm and intensity are as much an integrated part of speech prosody as F₀ contour patterns. Therefore, we conclude that in order to construct a working prosody model, all three acoustic correlates should be considered in relation to prosody organization. The conclusion is also directly applicable to TTS to improve output naturalness.