Audio Effect for Highlighting Speaker's Voice Corrupted by Background Noise on Portable Digital Imaging Devices

In this paper, an audio effect (AE) algorithm is proposed which can be applied to portable digital imaging devices to enjoy video contents effectively. The proposed AE algorithm enhances speech signals corrupted by background noise in audio content based on audio content classification (ACC) and the signal-to-noise ratio (SNR) estimation in order to highlight speaker's voice. The ACC classifies each short segment of audio content as speech, non-speech, or mixed signal by using the parameters such as signal energy, sub-band energy, and residual signal energy obtained from the linear prediction analysis. Then, we adaptively scale the signals according to the classification and the estimated SNR. To show the effectiveness of the proposed AE algorithm, we perform an informal listening test between the original audio contents and their processed versions by the proposed AE algorithm. Consequently, it is shown that the proposed AE algorithm significantly improves audio quality.