Abstract
Since communication is essential to all people, speech disorders cause great distress. In this study, we proposed a speech support system for articulation disorders, resulting from dysfunctions of the articulatory organ that make it difficult for the individual to vocalize accurately. We adapt the method used in a previously developed speech support system to articulation disorders. In our previous studies, we focused on phonation disorders caused by the loss of vocal cords through pharyngeal cancer. However, since the difference between articulation and phonation disorders is merely the position of the impairment, the proposed system could also be applied to articulation disorders. This is accomplished by incorporating the LPC (linear predictive coefficient) of articulation estimated from a healthy person into the BCS (body-conducted speech) of the individual with the disorder. For ease of use in daily life, we have implemented this system using portable devices like smart phones. In this study, we focused on two aspects. First, we confirmed the validity of this system for articulation disorders. Concretely, we made the BCS clearer using LPC, and evaluated the articulation quality using discrete word recognition. Second, we investigated the effect of speaker adaptation to the articulated speech. In order to implement our system on portable devices to facilitate practical use, we conducted recognition using speaker adaptation for the articulated speech. This resulted in a highly improved recognition rate. Therefore, we confirmed that our system is applicable to articulation disorders, and has acceptable recognition precision using speaker adaptation.
