A new frequency domain approach to time-scale expansion of audio signals

We present a new algorithm for time-scale expansion of audio signals that comprises: time interpolation, frequency-scale expansion and modification of a spectral representation of the signal. The algorithm relies on an accurate model of signal analysis and synthesis, and was constrained to a non-iterative modification of the magnitudes and the wrapped phases of the relevant sinusoidal components of the signal. The structure of the algorithm is described and its performance is illustrated. A few examples of time-expanded wideband speech can be found on the Internet.