Improved Time Delay Analysis/Synthesis for Parametric Stereo Audio Coding

The time delay for each partition is estimated as the slope of the phase difference between left and right channels Φ(k, i) = arg (X1(k, i)X ? 2 (k, i)) . Below the phase wrapping boundary the slope corresponds to phase delays, and to group delays above. Φ̂(k, i) = a1i or Φ̂(k, i) = a1i + a0 The linear regression is done in the weighted least squares sense, with weights chosen as a function of the magnitude of the crossspectrum between left and right. [ a1 a0 ]T = ( HWH )−1 HWy , Where H is a N × 2 matrix of frequency indices (first column) and ones (second column), and yi = Φ(k, i).