2021/01/30 by Protima Nomo Sudro, Rohan Kumar Das, Sudro, Protima Nomo +5
Biochemistry, Genetics and Molecular Biology · Computer Science · #Audio and Speech Processing (eess.AS) #Cleft Lip and Palate Research #FOS: Electrical engineering #Speech Recognition and Synthesis #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2102.00270
openalex publication_date 2021/01/30 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Cleft lip and palate (CLP) refer to a congenital craniofacial condition that\ncauses various speech-related disorders. As a result of structural and\nfunctional deformities, the affected subjects' speech intelligibility is\nsignificantly degraded, limiting the accessibility and usability of\nspeech-controlled devices. Towards addressing this problem, it is desirable to\nimprove the CLP speech intelligibility. Moreover, it would be useful during\nspeech therapy. In this study, the cycle-consistent adversarial network\n(CycleGAN) method is exploited for improving CLP speech intelligibility. The\nmodel is trained on native Kannada-speaking childrens' speech data. The\neffectiveness of the proposed approach is also measured using automatic speech\nrecognition performance. Further, subjective evaluation is performed, and those\nresults also confirm the intelligibility improvement in the enhanced speech\nover the original.\n