Alveolar fricative consonants detection with easily interpretable feature for speech training
Kouki Furuya, Arata Kawamura, Youji Iiguni · 2019
In this study, we conducted feature extraction for speech training by hearing-loss people alone. We targeted alveolar frictional consonant (/ s, z/), which is particularly error prone, and divided its features into three parts: fricativeness, articulation point, and voicing. The indicators created by these features are easily interpretable and help correct incorrect pronunciation.