UNIVERSITY OF BUCHAREST
FACULTY OF PHYSICS

Guest
2024-11-23 17:57
 HOME     CONFERENCES     SEARCH            LOGIN     NEW USER     IMAGES   


Conference: Bucharest University Faculty of Physics 2016 Meeting


Section: Theoretical Physics and Applied Mathematics


Title:
Statistical indicators of phonetical balance for Romanian utterances


Authors:
A. COCIOCEANU(1), T. IVANOAICA(1,2), A.I. NICOLIN (1,2), M.C. RAPORTARE (1)


Affiliation:
1) Department of Computational Physics and Information Technologies, Horia Hulubei National Institute for Physics and Nuclear Engineering, Reactorului 30, Magurele, Romania

2) University POLITEHNICA of Bucharest, Splaiul Independentei 313, Bucharest, Romania


E-mail
nicolin@nicolin.info


Keywords:


Abstract:
Given the surge of voice biometrics we investigated by means of computer-based statistical analyses the concept of phonetical balance for Romanian utterances. Our main result is that the standard distribution of phonems offers and incomplete description of the phonetics of the language and that more detailed statistical indicators are needed. For this purpose we introduce a simple indicator that measures the local vowel-consonant correlations and analyze the distribution of consonant clusters for Romanian words. We show that these two indicators complement the existing distribution of phonems [1,2] and represent the starting point for future investigations which use indicators with long-range correlations.


References:

[1] Stanescu (Pasca), M., Cucu, H., Buzo, A., Burileanu, C.: ASR for Low-Resourced Languages: Building a Phonetically Balanced Romanian Speech Corpus, 20th Eu- ropean Signal Processing Conference (EUSIPCO 2012), EURASIP 2012, ISSN 2076-1465.

[2] Stanescu (Pasca), M., Buzo, A., Cucu, H., Burileanu, C.: Statistical Phonetic Anal- ysis of the Romanian Language for Speech Recognition and Synthesis Tasks, 54th International Symposium ELMAR 2012.

Acknowledgement:
This work was supported by the H2020 SpeechXRays H2020 project and by ANCSI through project PN 16420202/2016. The authors thank Dan Caragea for insightful discussions.