Phonemic Restoration Based on the Movement Continuity of Articulation

Abstract

Phonemic restoration describes one of human capabilities that can retrieve the defective speech signal after adding a certain noise. It is believed that the movement continuity of articulation is one of main factors for realizing the phonemic restoration. This paper proposes an effective method based on this consideration to retrieve missing speech signal and makes the relevant hypothesis verified to some degree. For the proposed method, the mapping relationship between acoustic and articulatory features is established based on deep neural network (DNN), where a hierarchical DNN architecture with bottleneck feature is realized to improve the performance for acoustic-to-articulatory inversion, then missing articulatory feature obtained from missing speech signal is restored with cubic spline function. 25 sentences are selected from the database MNGU0 and short durations of the sentences are replaced by zeros and/or noise for evaluating. Experimental results show that the proposed method can effectively improve perceptual evaluation of speech quality (PESQ) of the speech with missing signal. And these experimental results provide preliminary experimental clues for verifying the first hypothesis of phonemic restoration—coarticulation.