Maveba 2021, Evaluating the Nasalisation of the Singing Voice

This study presents a vocal analysis prototype tool for the quantification of the nasality characteristic in singing. The tool is part of a larger, under development, software suite for the singing voice, which is aimed towards the analysis of both qualitative and quantitative vocal characteristics from specific audio samples. The tool examines the nasality characteristic by assessing formant central frequencies and bandwidths. In order to determine the most relevant acoustic parameters to be extracted and analysed by the tool, a case study experiment was conducted on two professional singers, singing in various degrees of willingly nasal voice. The audio samples recorded through this process were submitted to perceptual evaluation and acoustic analysis. Statistic analysis of the results demonstrate higher correlation coefficients between the nasality rating and the second formant frequency, followed by the first formant bandwidth. This prototype version relies on a regression model based on the above statistics.