Abstarct: The speaker recognition has been one of the interesting issues in signal and speech processing over the last few decades. A lot of efforts have been applied to improve the efficiency of speaker recognition system. Feature extraction is one of the main part of a speaker recognition system which can improve the performance of system. There are some methods in order to choose better feature, but clustering is used in this project to find feature vectors that have the most similarity. Therefore, these similar vectors specify the most properties in vocal tract of each person that they help us to build acoustic model of each person or obtain better decision boundary between different individuals. We’ve purposed two methods that ICA is used in both and then use k-mean and PSO algorithm have been used in order to compare ICA’s performance. Finally, two purposed methods using ICA have been compare to SVM and ELM optimaized in recognition rate and period of time used databaxse. Experimental results show that recognition accuracy has been improved in perposed methods by using ICA. In the project, we use English Language Speech Databaxse for Speaker Recognition as a databaxse. Therefore, our system is text independent speaker identification.