Speech signals with Russian vowels were recorded simultaneously by two microphones. The first microphone was located in the larynx near the vocal chords and the second one was outside a mouth near the lips. A signal in the inner microphone is formed by the vocal chords and contains a weak reverberation echo. It is clipped to a half of the energy because of the sensitivity restrictions on the microphone. A new mathematical algorithm is proposed for restoration of the clipped signal part. The restored signal sounds better. The restored spectra of vowels contain the first formant that cannot be explained by a backward reverberation. A transfer function of the vocal tract is estimated by comparing spectra of the signals from the input and output microphones.
CITATION STYLE
Barabanov, A., Evdokimova, V., & Skrelin, P. (2015). Estimation of vowel spectra near vocal chords with restoration of a clipped speech signal. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 9319, pp. 209–216). Springer Verlag. https://doi.org/10.1007/978-3-319-23132-7_26
Mendeley helps you to discover research relevant for your work.