End-to-End Acoustic Feedback in Language Learning for Correcting Devoiced French Final-Fricatives

Abstract

This work aims at providing an end-to-end acoustic feedback framework to help learners of French to pronounce voiced fricatives. A classifier ensemble detects voiced/unvoiced utterances, then a correction method is proposed to improve the perception and production of voiced fricatives in a word-final position. Realizations of voiced fricatives contained in French sentences uttered by French and German speakers were analyzed to find out the deviations between the acoustic cues realized by the two groups of speakers. The correction method consists in substituting the erroneous devoiced fricative by TD-PSOLA concatenative synthesis that uses exemplars of voiced fricatives chosen from a French speaker corpus. To achieve a seamless concatenation the energy of the replacement fricative was adjusted with respect to the energy levels of the learner's and French speaker's preceding vowels. Finally, a perception experiment with the corrected stimuli has been carried out with French native speakers to check the appropriateness of the fricative revoicing. The results showed that the proposed revoicing strategy proved to be very efficient and can be used as an acoustic feedback.