Time-scale modification of audio signals using multi-relative onset time estimations in sinusoidal transform coding

The traditional sinusoidal transform coder (STC) was originally developed for analysis, synthesis, and modification of speech and other non-polyphonic audio signals. In this paper, a novel method of time-scale modification for polyphonic and multi-pitch audio signals based on STC is proposed. The proposed method does not require a pitch estimation, and as such, it enables STC based algorithms to perform modifications on polyphonic audio signals. The frequency jitter artifacts in the traditional STC are mostly due to the inaccurate onset time estimates measured by pitch periods. The proposed method eliminates the frequency jitter artifacts significantly by using multi-onset time estimations.