Very often in signal analysis it is necessary to determine certain points in the signal waveform as anchor points for the processing. This paper proposes a strategy to find automatically the voice onset for initial vowels, the burst point for voiceless and voiced stops and the release point for nasals. The presented approach exploits only the information inherent in the waveform of the signal by imitating the human skill of abstraction. The procedure is divided into the following steps 1. data compression by a) coding the original signal sampled at 16 kHz by its local extrema and their position in time b) eliminating the silent parts of the signal c) signal smoothing 2. classification of the syllables by their type 3. special treatment for each type of syllable