phoneme recognition can be viewed as classifying multivariate observation. Breeder Genetic Algorithm (BGA) and Genetic Algorithm (GA) approach the decision problem using two complementary models. For technical speech recognition systems as well as for humans it has been shown that the combination of acoustic and optical information can enhance speech recognition performance. In this paper we systematically investigate the phoneme recognition problem, where phones are a useful representation because words can easily be-written as phones using a lexicon. The propose models depend on three stages: the first stage is preprocessing of speech by token speech into phonemes, the second stage includes selection phonological features over phones (i.e. cepstral coefficients), the final stage includes using the Evoluationary Algorithm (BGA & GA) to clustering the dataset and recognition each phonemes