Automatic Voice Disorder Detection from a Practical Perspective

Abstract

Voice disorders, such as dysphonia, are common and often go untreated until they become severe. Assisting the detection of voice disorders could facilitate early diagnosis and subsequent treatment. In this study, we address the practical aspects of automatic voice disorders detection (AVDD). Data annotated for voice disorders is usually scarce due to challenges involved in data collection and annotation of such data. However, some relatively large datasets are available for a reduced number of domains. In this context, we propose using a combination of out- of-domain and in-domain data for training a deep neural network-based AVDD system and offer guidance on the minimum amount of in-domain data required to achieve acceptable performance. Further, we propose the use of a cost-based metric, the normalized expected cost (EC), to evaluate performance of AVDD systems in a way that closely reflects the needs of the application. As an added benefit, optimal decisions for the EC can be made in a principled way given by Bayes decision theory. Finally, we argue that for medical applications like AVDD, the categorical decisions need to be accompanied by interpretable scores that reflect the confidence of the system. Here, we show that adding a calibration stage-trained with a small amount of in-domain data can improve these models and support professionals in their decision-making.