Modeling Inhalation in Voice Activity Detection
Josafá de Jesus Aguiar Pontes · 2019
The present work investigates the detection of voice activity in acoustic signals of human speech that occur in low noise environments. It proposes a method capable of differentiating between speech and pause events in an acoustic signal that includes inhalations. For this purpose, it uses components, which are based on Support Vector Machine (SVM) classifiers, specialized in detection of both silence and inhalation. In this sense, it shows that the detection of inhalations allows for improving the accurate prediction rate of speech and pause events in a VAD system, reaching a result of 71.8%.