The development of voice disorder evaluation system based on dysphonia severity index

摘要

Voice disorder is dramatically increasing due to the unhealthy social habits such as smoking and alcohol consumption, voice abuse, and the most importantly the lack of awareness among the general public and from the health care provider. Objective non-invasive multiparameter voice assessment is seen as a way to improve the voice rehabilitation process by allowing home care at own responsibility. The purpose of the research is to develop an automatic voice diagnostic system based on objective non-invasive multiparameter method known as Dysphonia Severity Index (DSI). DSI consists of four parameters which are the highest pitch, jitter percentage, lowest intensity, and maximum phonation time. They are combined into a linear regression equation that will give values from -5 to +5 indicating severely dysphonic voice or normal voice respectively. The proposed system is named as Automatic Dysphonia Evaluation System (ADES). It integrates a new proposed pitch detection algorithm (PDA), start/end point detection algorithm, jitter equation, and intensity equation to obtain the four DSI parameters allowing the system to be used by patient at home to monitor their voices. The proposed PDA was proven more accurate by having no error detected for normal voice while only one pathological voice was detected with doubling error. The modified start/end point detection algorithm is proven better with silence detection error rate of 0.0752. ADES was tested with KayPENTAX voice database and had 55.6054% sensitivity and 50% specificity when -6.7249 is used as the cutoff value. Different sets of database consisted of trained and untrained vocalists, and also teachers and non-teachers were also used to evaluate ADES’ performance. The results of ADES show that it is able to get the DSI values for different voices from different types of groups.