In this contribution a parameter is introduced which is commonly used in human phoniatrics - the Harmonics to Noise Ratio (HNR). It is a ratio of the acoustic energy of the harmonic components to that of the noise. Barks of domestic dogs are short, loud and broad band utterances. The rationale for our method is that the acoustic wave of a barking call is characterised by the ratio of harmonic (i.e. periodic oscillation of the vocal folds) and noise components (i.e. air turbulence in the vocal tract as well as aperiodic vibration pattern of the vocal folds). The HNR is estimated as follows: a 50 ms - tile segment from the middle of the call is Fourier transformed (sampling frequency 30 kHz, 512 point Hanning window). A ten point moving average curve was calculated from the spectrum. The maximum difference between the spectrum and the moving average curve is used as HNR. 50 barks from each of ten adult healthy dogs (breed: dachshound) uttered during mild threat were considered. In four abnormal sounding dogs (to the human ear) the HNR was also calculated. Three out of those four were out of the range of the normal sounding dogs: although there was a considerable overlap. Test persons asked for ranking dog barks along a scale from 'very noisy' to 'very harmonic' showed good coincidence with the HNR calculated for those calls. This exemplifies that the relatively simple HNR is valuable in animal bioacoustics (supported by the Land Berlin, NAFoG).
Abstract: