Little research has been introduced for giving feedback to consonant sounds so far because of a reliability problem in the performance of speech recognition for consonant sounds. On the other hand, vowel sounds has been focused on for many CALL systems. To solve the problem, we have adopted a novel feedback method based on the classification of the place and manner of articulation. Actually, as we perform the categorical recognition for both articulations instead of consonant phonemes themselves, we can afford a stable and reliable feedback to the learners (It will be mentioned in more details in section 4.5). The classified results for the pronunciation errors are processed in the feedback module based on these categorization, and are used to instruct the learners on their articulatory configuration. Namely, for the purpose of automatic instruction, we have classified Japanese consonants into the categorical articulation according to a few well-recognized features: the place of articulation and the manner of articulation as described in Table 2.1.
