Voice Feature Analysis for Early Detection of Voice Disability in Children

Abstract

Voice disability is one of the common disabilities encountered by human being. Around 1.2% of the World’s population have been diagnosed with some form of voice disability. Painful endoscopic procedures are commonly practiced to detect this disability. In the recent years, researchers are trying to discover alternatives to avoid these painful procedures. Detecting voice disability by using voice sample analysis is one of them. However, detecting voice disability in children by using this method is a very challenging task because of their immature voice generation system. There is always a chance of misdiagnosis. Hence, it is very imperative to choose appropriate signal processing techniques. In this paper, we investigate different signal processing techniques to detect voice disability in children. Based on the results, we conclude that spectrogram, wavelet and MFC (Mel Frequency Cepstral) are the three main techniques that can distinctly detect voice disability in children.