Automatic band segmentation based on spatio-frequency processing using MUSIC algorithm

Abstract

Automatic band segmentation (channel isolation or information channel identification), is a key problem in automatic classification (recognition) of radiocommunication signals. All automatic radiocommunication signal classification methods published in the literature (Liedtke, 1984 and Domingues et al, 1991), are sensitive to the channel isolation or mistuning of the center frequency, or mismatch of the bandwidth and crosstalk from the adjacent frequency channels. A new approach to automatic band segmentation or channel isolation is proposed in this paper which is based on joint estimation of spatial and spectral parameters of radio signals which are superimposed in a selected radio frequency subband which is based on the MUSIC algorithm. It provides a general solution to the channel isolation problem even in the situation where many active emitters are overlapped both in time and frequency and in the case of worse propagation conditions such as multipath.