Summary

This 2007 blog describes an FFT-based technique to estimate and correct time delay and scaling between two signals, with a practical Matlab/Octave implementation (fitSignal). It explains phase- and correlation-based approaches, demonstrates how to get sub-sample accuracy, and shows how simple preprocessing (windowing, zero-padding) and interpolation improve robustness.

Key Takeaways

• Use FFT-based cross-spectral phase to estimate relative delay between two signals efficiently.
• Apply phase unwrapping and interpolation to achieve sub-sample (fractional) delay estimates.
• Implement the provided fitSignal Matlab/Octave routine to compute delay and scaling and to realign signals.
• Mitigate noise and bias by using windowing, zero-padding, and optionally GCC-PHAT weighting for robustness.

Who Should Read This

Intermediate DSP engineers, researchers, or graduate students working on audio, radar or communications who need a concise, practical FFT-based method and Matlab code for time-delay estimation and signal alignment.

TimelessIntermediate

Topics