Nowadays, some speech signal processing technologies such as speech encoding, synthesis, recognition and so on, are applied in our daily lives. But to some extent the existence of noise baffles the further application of these technologies. To solve this problem, speech enhancement and separation emerge as the times require.In this paper, the research is mainly about speech separation. The most two popular algorithms are Computational Auditory Scene Analysis (CASA) and Blind Source Separation. This paper proposes an algorithm called repeating pitch adding (subtracting) and another algorithm which is an improvement of Blind Signal Separation. The purpose of these new methods is to reduce the quantity of operation, shorten the operating time and assure that it can separate overlapping speech effectively. The experiments verify that these two algorithms can attain this purpose. Therefore, the two methods proposed in this paper are more suitable for the daily application. These algorithms are effective when they are used to separate overlapping speeches of two voices, but the result of the experiment in the noisy environment is not very good. So we need to do more research to solve this problem. |