Speech analysis for identification and recognition purposes is a demanding task, especially to find recognition parameters for some consonants. In this paper, we show the initial analysis results of speech corpus, sampled with two sampling rates, 22050 and 8000 Hz, using Higuchi fractal dimension. The study shows that it might be possible to use fractal dimension value for classification of for example plosives /k/, /p/ and /t/. However, the analysis seems to be very sensitive to the sampling frequency in speech analysis.