Mispronunciation detection leveraging maximum performance criterion training of acoustic models and decision functions

Mispronunciation detection is part and parcel of a computer assisted pronunciation training (CAPT) system, facilitating second-language (L2) learners to pinpoint erroneous pronunciations in a given utterance so as to improve their spoken proficiency. This paper presents a continuation of such a general line of research and the major contributions are twofold. First, we present an effective training approach that estimates the deep neural network based acoustic models involved in the mispronunciation detection process by optimizing an objective directly linked to the ultimate evaluation metric. Second, along the same vein, two disparate logistic sigmoid based decision functions with either phone- or senone-dependent parameterization are also inferred and used for enhanced mispronunciation detection. A series of experiments on a Mandarin mispronunciation detection task seem to show the performance merits of the proposed method.