ISCA Archive Interspeech 2022
ISCA Archive Interspeech 2022

Phase Vocoder For Time Stretch Based On Center Frequency Estimation

Donghyeon Kim, Bowon Lee

In this paper, we propose a phase correction algorithm for phase vocoders based on center frequency estimation. Phase shift estimation in the phase vocoder for time stretch is one of the important problems, for which there exists no consistent solution. In many cases, a simple method consisting of STFT, phase correction, and ISTFT cannot produce high-quality results, so methods such as using multiple FFT sizes and sine wave synthesis through oscillator bank are proposed, which unfortunately is computationally expensive. The proposed method uses a phase correction algorithm that can generate high-quality results while enabling a high degree of parallelism for faster execution. It uses a simple structure that includes only STFT, phase correction, and ISTFT as essential elements. The phase correction algorithm uses peak phase-locking and a method to find an appropriate dominant peak frequency, and it requires only a single sized FFT and has the advantage that it can be easily applied to various applications thanks to its structural similarity to the classical phase vocoder.