Performance comparison on vocal cords disordered voice discrimination via machine learning methods*

Abstract

This paper studies how to improve the identification rate of laryngeal disability speech data by convolutional neural network (CNN) and machine learning ensemble learning methods. In general, the number of laryngeal dysfunction speech data is small, so even if identifiers are constructed by statistical methods, the phenomenon caused by overfitting depending on the training method can lead to a decrease the identification rate when exposed to external data. In this work, we try to combine results derived from CNN models and machine learning models with various accuracy in a multi-voting manner to ensure improved classification efficiency compared to the original trained models. The Pusan National University Hospital (PNUH) dataset was used to train and validate algorithms. The dataset contains normal voice and voice data of benign and malignant tumors. In the experiment, an attempt was made to distinguish between normal and benign tumors and malignant tumors. As a result of the experiment, the random forest method was found to be the best ensemble method and showed an identification rate of 85%.

AI-based stuttering automatic classification method: Using a convolutional neural network*

“…터 획득이 가능해지면서 CNN(convolutional neural network)을 적용한 장애 음성 식별 연구사례들이 보고되었다(cf. Bhushan et al, 2021;Fang et al, 2019;Jo et al, 2022;Prabhu & Seliya, 2022) (1)…”

“…일반적으로 말더듬 평가는 빈도나 비율에 기 반한 청지각적 판단을 필수적으로 고려하여 수행된다 (Tichenor et al, 2022). 하지만 평가 수행에 있어 긴 소요시간과 평가자 간 신뢰도(reliability) 문제가 야기될 수 있다 (Kully & Boberg, 1988;Yaruss, 1997) (Jo et al, 2022).…”