A novel convolutional neural network based dysphonic voice detection algorithm using chromagram

<span>This paper presents a convolutional neural network (CNN) based non-invasive pathological voice detection algorithm using signal processing approach. The proposed extracts an acoustic feature, called chromagram, from samples and applies this feature to the input of CNN for classification. main advantage chromagram is that it can mimic way humans perceive pitch in sounds hence be considered useful detect dysphonic voices, as generated varies depending on conditions. simulation results show classification accuracy 85% achieved with chromagram. A comparison performances those other related works also presented.</span>