An Evolving Model of Voice Disorder Detection using Deep Belief Network

摘要

In recent years, automatic diagnose of larynx pathological voice disorders are a challenging task in the medical filed. The researchers started focusing on working with voice signals to discover voice disorder related diseases. Machine learning plays a vital role in automatic detection of voice disorder using spectral information of recorded voice. Among several approaches deep playing has been in a prominent place for achieving significant results in the voice recognition field, where there has been less research work in the field of pathological voice detection. This paper introduces the deep belief network for discovering healthy and unhealthy voice detection. The stack of Restricted Boltzmann Machine is used to pretrain the deep neural networks. Simulation analysis is done to prove the proficiency of the deep belief networkbased voice disorder detection using the real data from the Saarbrucken Voice database..