Spectral analysis is an important technique that is often utilised in speech processing. In the case of speech synthesis it is often used for the extraction of formant information, from natural speech, which is then used to construct a data base of the synthesisers drive parameters for subsequent synthesis. In the case of analysis the extracted frequency information is often used in its own right. Traditionally the Fast Fourier Transform has been the technique adopted to perform the required frequency extraction. The wavelet transform has recently gained favour for carrying out frequency analysis due to its ability to localise a signal in both frequency and time. This paper describes the investigations carried out into the use of the wavelet transform for formant extraction. The resynthesis of the natural speech using a parallel formant speech synthesizer driven using the extracted formant information. Comparisons are drawn as to the effectiveness of the technique when compared to that of a group delay minimum phase FFT technique.