New wavelet transform normalization to remove frequency bias

A new wavelet transform normalization procedure is proposed for the construction of a weighted bank of correlation filters. The standard normalization results in lower input frequencies producing larger wavelet transform magnitudes for equal-amplitude frequencies, while the new normalization produces equal responses as desired. This is illustrated with examples of Gibbs overshooting phenomenon and a cocktail party effect. A derivation is given to show that an inverse transform still exists when using the new normalization.