Interpolated DFT algorithms with zero padding for classic windows

• A new interpolation algorithm is proposed.
• It can be applied to most classic window.
• The algorithm was more robust than traditional algorithms in the presence of noise.
• It is insensitive to the incorrect polarity estimation problem.

It is well known that accurate frequency estimation of multi-frequency signals contaminated with additive noise is a common problem encountered in a large number of engineering practice scenarios. This paper proposes a new frequency estimator based on using zero-padding and main-lobe fitting techniques. The primary innovation of this new algorithm is that it can be applied to most classic window functions, including adjustable windows. Systematic errors for various windows are studied and algorithm stability with respect to white Gaussian noise is investigated. In addition, a comparative study demonstrates that the proposed algorithm is more robust against additive noise than traditional algorithms because of its insensitivity to the incorrect polarity estimation and an intrinsic ability for partially canceling noise influence. Featured with straightforward operation, sufficient accuracy, strong compatibility, as well as robustness towards additive noise, the proposed approach is an appropriate choice for frequency estimation in spectral analysis.