A user voice reduction algorithm based on binaural signal separation for portable digital imaging devices

In this paper, a user voice reduction algorithm for portable digital imaging devices is proposed based on a binaural signal separation approach in order to improve the naturalness of user-generated video contents. The proposed algorithm first estimates the interaural time differences (ITDs) from binaural signals recorded by the microphones equipped on a device. Then, the estimated ITDs are used to obtain the time-frequency domain masking patterns of a user voice against an actual subject sound of video content. Finally, the user voice recorded in video content can be reduced by applying the mask patterns to the binaural signals. In order to demonstrate the effectiveness of the proposed algorithm, the proposed algorithm is implemented on a portable digital imaging device having a clock speed of 600 MHz. It is shown from the performance evaluation by measuring a sound pressure level that the proposed algorithm reduces user voice by around 10 dB.