Blind sources separation (BSS) arises in a variety of fields in speech processing such as speech enhancement, speakers diarization and identification. Generally, methods for BSS consider several observations of the same recording. Single microphone analysis is the worst underdetermined case, but, it's also the more realistic one. In our approach, the autoregressive structure (short term prediction) and the periodic signature (long term prediction) of voiced speech signal are jointly modeled. The filters parameters are extracted using a combined version of the EM-Algorithm with the Rauch-Tung-Striebel optimal smoother while the fixed-lag Kalman smoother algorithm is used for the initialization.