A robust and accurate cross-correlation-based fundamental frequency(F₀) determination method for the improved analysis of infantcries

摘要

The extraction and analysis of the fundamental frequency (F₀) of infant vocalizations has been the focus of a number ofefforts over the past thirty years. It is thought that this parameter isan important information channel from which information regarding thestate of an infant can be determined. To date, research groups workingto extract the vocal fundamental frequency of infant cries have beenlimited in the resolution and granularity of the F₀extraction methods used for adult speech, which, typically are not wellsuited for infant cries. In general, the fundamental frequency range ofadult speech is limited to values below 600 Hz, whereas for infantcries, F₀ can have a range of several kilo hertz, and besubject to rapid changes in certain cases. This paper presents a newmethod for accurately determining the F₀ of infant cries,which should be applicable to other vocalizations as well. The methodpresented uses the crosscorrelation of adjacent speech segments togenerate a three-dimensional plot called a crosscorrelogram. From thisplot, the fundamental frequency of an infant cry can easily beextracted, with increased precision