Tools for the analysis of acoustical and physiological characteristics of the singing voiceIngrid Bendl
Deriving acoustical, musical and physiological properties of the singing voice by analyzing recorded sound files is a good way to provide feedback for the singing student as well as valuable diagnostic information for the speech therapist. For this kind of analysis there exist a variety of public domain computer programs and computation packages which will be introduced and demonstrated.
A common framework with a graphical user interface will be presented which can be used to combine the strengths of several existing tools, to pre-process recorded audio files, to automate repetitive analysis procedures on file sets and to post-process obtained results statistically. The framework integrates segmentation, time domain analysis, frequency domain analysis and statistical analysis. It is based on well known tools like Praat, Wavesurfer, SNDAN, TAP, ProToo and R-project as well as on Matlab scripts like VoiceSauce (HNR, spectral tilt, formants) and Aparat (Inverse Filtering).
Resulting data are long time average spectra and their band energy ratios like α-factor, short time spectra and their time varying characteristics like loudness (rms), brilliancy (normalised spectral centroid), vibrato (pitch), formants (LPC) as well as physiological parameters represented by spectral tilt (H1-H2, H1-A1, H1- A2, H1-A3), HNR and other glottal parameters.