An Improved Method Using STFT for Separation of Speech Signals
C.Anna Palagan1, K.Parimala Geetha2, T.Leena3

1Dr. C. Anna Palagan, Associate Professor, Department of Electronics and Communication Engineering, Malla Reddy Engineering College, Maisammaguda, Hyderabad, Telangana, India.

2Dr. K. Parimala Geetha, Professor and Head, Department of Electronics and Communication Engineering, Ponjesly College of Engineering, Kanyakumari, Tamilnadu, India.

3T. Leena, Associate Professor, Department of Electrical and Electronics Engineering, Rajas Engineering College Tirunelveli, Tamilnadu, India.

Manuscript received on 05 March 2019 | Revised Manuscript received on 12 March 2019 | Manuscript Published on 20 March 2019 | PP: 291-294 | Volume-8 Issue- 4S2 March 2019 | Retrieval Number: D1S0064028419/2019©BEIESP

Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open-access article under the CC-BY-NC-ND license (

Abstract: The key purpose of this paper is to recuperate the intention module of speech mixed with interfering speech, and to advance the recognition accuracy. This is attained by the improved speech signals, which is designed for effectually separated the speech signal from the blind source separation by expending the Instantaneous Mixing Auto Regressive method and the maximum prospect function. The significant features present in the Instant Mixing of Auto Regressive is that it gets boosted the split-up of speech signals and thereby aiding us to perform a blind source fragmentation process in contemplation, the Signal and Interference Ratio rate progresses over 6 dB. By using Instantaneous Mixing Auto Regressive method(IMAR) it accomplished good signal and interference ratio along with direct and reverberation ratio even though the reverberation time was 0.3 sec only. In this research work, dual channel and single channel speech fragmentation and enhancement algorithms are discussed and the performances of the proposed algorithms are analyzed in detail based on the objective and subjective quality measures. For the experimental setup, we consider the 0.3 sec and 0.5 sec reverberation time.

Keywords: STFT, IMAR, Mixing Matrix, Separation Matrix, Prediction Matrix.
Scope of the Article: Communication