Table 3 Paradigm-wise (a)–(d) top ten acoustic features according to the ascending mean (given below the respective feature name outside the brackets) of cross-validation iteration-wise (mean of sub-models) feature ranks (given below the respective feature name within the brackets) for both vocalisation-wise and infant-wise evaluation scenario.

From: Automatic vocalisation-based detection of fragile X syndrome and Rett syndrome

  1. Mean ranks are rounded to two decimal places. A, amplitude; AD, atypical development/atypically developing; B, bandwidth; \(\text{f}_{o}\), fundamental frequency; \(\text{F}_{2,3}\), second and third vocal formant; FXS, fragile X syndrome; HNR, harmonics-to-noise ratio; idx, index; \(\text{L}_{eq}\), equivalent sound level; \({\text{MFCC}}_{1,2,3,4}\), first, second, third, and fourth Mel-frequency cepstral coefficient; pctl, percentile; pctlr, percentile range; RTT, Rett syndrome; \(\text{SD}_{\text {norm}}\), standard deviation normalised by the arithmetic mean; TD, typical development; URs, unvoiced regions; VRs, voiced regions; ♀, female; ♂, male; colour code for features: light grey = frequency-related feature, middle grey = spectral/cepstral feature, dark grey = energy/amplitude-related feature.
Back to article page