Blind estimation using higher-order cumulants

This paper presents a higher-order cumulant based blind estimation algorithm for estimating the channel matrix in a n-sensor-m-source system under a noisy environment. The output from an array of n spatially separated sensors is the only available information that can be used to derive the channel parameters. The algorithm requires neither the additive Gaussian noise to be white nor the number of sensors to be greater than the number of sources. We first evaluate the higher-order cumulant matrix from the sensor outputs, and the number of sources is then deduced from the rank of the matrix. Subsequently, by using the cumulant properties for independent sources, a permuted and scaled channel matrix is obtained. Simulation results illustrate the validity and potential applications of the method