作者: Shahrzad Shapoori , Saeid Sanei , Wenwu Wang
DOI: 10.1109/MLSP.2015.7324342
关键词:
摘要: Sparsity is known to be very beneficial in blind source separation (BSS). Even if data not sparse its current domain, it can modelled as linear combinations of atoms a chosen dictionary. The choice dictionary that sparsifies the important. In this paper partly pre-specified based on chirplet modelling various kinds real epileptic discharges, and learned using learning algorithm. which includes fixed variable (i.e. learned) part, incorporated into framework extract closest interest from mixtures. Experiments synthetic mixtures consisting discharges are used evaluate proposed method, results compared with traditional BSS