Speech Enhancement of Singing Support System for Lartngectomies
Hirotaka Nogami, Yoshihisa Nakatoh, Kenji Matsui · 2017
In this research, the purpose was to develop a singing support system for laryngectomies, we examined a singing sound source along with a sound quality improvement for the singing voice. For the singing sound source, the Bark LPC residual signal obtained by flattening the LPC residual signal with Bark spectrum was used. To improve the sound quality of singing voice, SS method which is one of speech enhancement methods was used. We carried out listening experiments for the two singing voices, “singing voice before processing” and “singing voice after processing”. We evaluated “Intelligibility”, “Noisy” and “Naturally”. As for the “Intelligibility” and “Noisy”, improvement effect was obtained for all phonemes. As for “Naturally”, there is a tendency to improve for /a/, /i/ and /o/, but there was almost no significance in /u/ and /e/.