Feature Maps of the Acoustic Spectrum of the Voice

J Voice. 2020 Jan;34(1):161.e1-161.e26. doi: 10.1016/j.jvoice.2018.08.014. Epub 2018 Sep 27.

Abstract

The change in the spectrum of sustained /a/ vowels was mapped over the voice range from low to high fundamental frequency and low to high sound pressure level (SPL), in the form of the so-called voice range profile (VRP). In each interval of one semitone and one decibel, narrowband spectra were averaged both within and across subjects. The subjects were groups of 7 male and 12 female singing students, as well as a group of 16 untrained female voices. For each individual and also for each group, pairs of VRP recordings were made, with stringent separation of the modal/chest and falsetto/head registers. Maps are presented of eight scalar metrics, each of which was chosen to quantify a particular feature of the voice spectrum, over fundamental frequency and SPL. Metrics 1 and 2 chart the role of the fundamental in relation to the rest of the spectrum. Metrics 3 and 4 are used to explore the role of resonances in relation to SPL. Metrics 5 and 6 address the distribution of high frequency energy, while metrics 7 and 8 seek to describe the distribution of energy at the low end of the voice spectrum. Several examples are observed of phenomena that are difficult to predict from linear source-filter theory, and of the voice source being less uniform over the voice range than is conventionally assumed. These include a high-frequency band-limiting at high SPL and an unexpected persistence of the second harmonic at low SPL. The two voice registers give rise to clearly different maps. Only a few effects of training were observed, in the low frequency end below 2 kHz. The results are of potential interest in voice analysis, voice synthesis and for new insights into the voice production mechanism.

Keywords: Harmonic-formant interaction; Level of the fundamental; Non-linear source-filter interaction; Spectrum balance (SB); Spectrum slope; Voice Range Profile (VRP); Voice Register.

MeSH terms

  • Acoustics*
  • Female
  • Humans
  • Male
  • Phonation*
  • Singing*
  • Sound Spectrography
  • Voice Quality*