A novel continuous speech segmentation method is described based on a measure of the Instantaneous Frequency function of the speech waveform. This function is employed for the estimation of a distance measure which in conjunction to established energy and duration threshold criteria gives the boundaries of speech phonemes, identified from vowels and non vowels as well as vowel and non vowel strings- This segmentation technique is more efficient computationally than established segmentation methods and its accuracy and performance is comparable to the best results obtained by the other methods. Results of application of 4 other methods are also given, when they were applied on a common Greek and English Language, continuous speech data base.