Blind separation of speech mixtures via time-frequency masking
IEEE Transactions on Signal Processing
Hi-index | 0.00 |
This paper proposes an optical access system that enables voice communication services to continue when a power failure occurs. In particular, we propose applying a digital signal processing technique namely blind speech separation (BSS) to an optical multiple access system for demultiplexing randomly mixture signals received from an upstream user. We present a brief overview and the target of proposed system and describe the key technique, namely the BSS procedure. We then report system feasibility studies for the first step of a numerical simulation (assuming two wavelengths and three users) and experimental tests (assuming two wavelengths and two users) using actual sample voice signals. We confirmed that separation was realized by using a subjective assessment that consisted of listening to obtained signals.