A eigenspace-based source number estimation was presented.It projects the estimated covariance matrix of array signal into signal eigen-subspace and noise eigen-subspace, respectively.Using the orthogonality between signal eigen-subspace and noise eigen-subspace, it is easy to differentiate the contribution of signal and noise by using the criterion value,or the magnitude of projection.Like the Direction-of-Arrival（DOA） estimation algorithm,the estimator uses the eigenvalue decomposition of covariance matrix with order M×M,where M is the number of elements,and hence can save much computational burden.To reduce more computational burden,the estimation can be implemented by the decomposition in the real-value space.Computer simulation demonstrates the distribution of criterion value and the performance of the estimation method.The estimation method was also tested with the sonar data,which shows good performances.
