Abstract
Source separation from instantaneous combinations of signal, so called observations, is a challenging problem in signal processing. Blind source separation (BSS) means that there is no information about the mixing matrix and observations. Blind source separation methods resort to very weak hypothesis concerning the source signals, as well as the mixing matrix. However, people often have priors on signals. A natural idea is to add these priors in the model, for simplifying or improving the separation methods. In this paper semi blind source separation (SBSS) based on least mean squares (LMS) and least squares (LS) was proposed to get the individual speech. LMS method was used to reduce noise traditionally, but in this paper we use it to separate mixed signals with some priors that will be explained in the introduction. Least squares method can also get better result in the condition of two speech signals mixed linearly with some prior knowledge. Meanwhile, in order to reduce noise, the SBSS method and the beamforming (BF) technology were combined to get better separation result in noisy environment. The result of simulation signals confirmed the feasibility of the method and it can be widely used in smart home.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call