The dynamic effect of speaking fast on speech prosody

Speaking fast causes several changes in speech prosody. In addition, it can be associated with a decrease in speech intelligibility. In this study, prosodic changes in fast speech were investigated using common prosodic measurements and syllabic prosody index (SPI), a novel prominence measure that combines f0, energy and duration features. Dynamic changes in long-term prosodic prominence were investigated using functional data analysis (FDA), in which the SPI is transformed into a functional form. The possibly decreasing effect of speaking fast on speech intelligibility was evaluated using automatic speech recognition. Phonetic analyses of syllabic units showed that speaking fast decreases duration, f0 and SPI, and increases articulation rate and proportional acoustic energy in the frequency range of 0–1 kHz. FDA supported the aforementioned results by revealing dynamically decreased overall prominence in fast speech. Furthermore, in comparison to regular speech, speech intelligibility was found to be significantly lower in fast speech: word error rate (WER) for regular speech was 0.27, whereas for fast speech it was 0.86.