Noise-resistant feature extraction using 2D techniques

The aim of this paper is to describe an algorithm for noise-resistant vowel identification in speech signals. The algorithm uses two-dimensional techniques of speech signal processing and with them automatically detects and identifies vowels in very noisy speech signals. Differentiating between vowels and noise is based on morphological analysis of periodical voiced parts of the speech signal and the noise. The wavelet transform of the noisy speech signal is used to convert the one-dimensional signal into a signal of two variables, which is further processed to increase the efficiency of vowel identification.