The traditional Hilbert Transform （HT） used in extracting and analyzing the envelop of speech signal suffers from some drawbacks

for example

unexpected blurs will appear in the extracted envelops which cause a lot of difficulties in the postprocessing of the speech signal.To overcome these draw backs a novel approach CAWT （Complex Analytical Wavelet Transform） for extracting and analyzing speech signal envelops is proposed in the present paper.The CAWT algorithm together with some simulation results are given which show many advantages over the HT extraction method.It is expected that the CAWT will be helpful in the Chinese speech recognition.