In this project, our aim is to examine new methods to find acoustic units in speech signal. We've started by studying literatures on the topic of automatic derivation of acoustic units for speech recognition. Beside that, we study some candidate methods like Particle Swarm Optimization (PSO). In Automatic Acoustic Unit Classification, the goal is to generate sub-word set(or phone set) and acoustic models optimally, and is generally a complex problem by nature because it involves joint optimization of some inter-related variables. If our studies show a possibilty of the usage of these methods in Automatic Acoustic Unit Classification problem then we proceed to the next stage, and implement and compare it with other techniques. This web page is the home page of this project, and all results, publications, reports, programs would be posted here. I you have any suggestion or question please feel free to contact us.