Spectral features in the singing voice evaluation using simulated imperfections

Abstract

Abstract The singing voice is an important element of cultural heritage. Apart from a few cases of people with innate talent and skills, its development takes place in the long-term learning process. The main motivation of the research presented in this paper is to compare the spectral characteristics of the human voice of professional singers and teachers based on the performance of selected singing fragments. The concept consists in performing selected vocal exercises, considering all components of best vocal practice and comparing them to performances in which the selected elements will be intentionally omitted or included to a lesser extent. The complexity of the problem lies in the interdisciplinary nature of the project and requires consideration of many areas in the field of acoustics, art, education, signal processing, etc. The selected subjective parameters of singing voice are presented, then the assumptions made, audio database construction and its content are described. Next, samples of graphic representations of recordings in form of modified spectrogram are shown. Then selected fragments of vocal exercises are analysed in terms of mutual relationship of spectral features. At the end, a summary and directions for further work are presented.