A Time-Scale Modification-Based Voice Changing Method with Seamless Switching and Its Real-Time Implementation on Digital Imaging Devices

Abstract

In this paper, we propose a voice changing method for providing a seamless switchable function with a low computational complexity for digital imaging devices. The proposed method combines a waveform similarity overlap-and-add (WSOLA) algorithm with a sampling rate changing technique that operates in the time domain. In addition, it includes a noise technique in the region where the voice changing switching mode changes from on to off, and vice versa. The performance of the proposed method is compared with that of a conventional one in terms of the speaker authentication rate, processing time, and speech quality. It is shown from the experiments that the proposed voice changing method provides a speaker authentication rate of 6% and gives a relative complexity reduction of 84.5% in a resource-constrained device with an ARM processor. Also it is shown that audio contents processed by the proposed method are more preferred than those by the conventional method.