Post-processing of time-frequency representations in instantaneous frequency estimation based on ant colony optimization

- Instantaneous frequency (IF) estimation of signals embedded in high noise is addresed.
- Ant Colony optimization method for post-processing the Wigner distribution (WD) is proposed.
- Besides WD, the proposed technique is also applicable on Short-time Fourier transform, Wavelet transform, S-method and L-WD.
- Cross-terms in the WD of multicomponent signals are reduced by the proposed algorithm.

Instantaneous frequency (IF) estimation of non-stationary signals embedded in high noise is addressed. When present, high noise represents a dominant error source in the IF estimation. Additive Gaussian noise with variance proportional to the signal power is assumed. An estimation approach based on the ant colony optimization (ACO) and time-frequency (TF) analysis is proposed. The ACO algorithm is adapted for the IF estimation starting from the Wigner distribution (WD) of the considered signal. The proposed technique is also applicable to numerous other representations, without any change in the parameter setup. This method surpasses the influence of high noise in the IF estimation; for that purpose, we have designed a pheromone deposition gradient and a mechanism for the variation of the agents' population size. The introduced approach improves the fast-varying IF estimation accuracy, overcoming known issues in the state-of-the-art algorithms dealing with high noise. The basic principles of the proposed method are illustrated and performance validated through numerical examples.