Performance Evaluation of Speech Enhancement of Single Acoustic Signal Referring to Image Information

Abstract

This paper presents performance evaluation of speech enhancement of an acoustic signal referring to image information. In recent years, we often see smart phones and tablets in everyday life. Smart phones and tablets have a camera and a microphone. Although many speech enhancement techniques use only the audio information from microphones, it is considered that visual information from a camera is also helpful to enhance the speech. Although the frame rate of the normal movie is very low compared to the sampling frequency of the audio signal, it is useful if it is adequately used. We investigated the performance depending on the parameters and confirmed its effectiveness through the experiments.