Abstract: To overcome the problem of signal-subspace dimension expanding, three different improved methods for wideband signal-subspace spatial-spectrum estimation are presented. The first method, called the signal-subspace dimension estimation method, estimates signal-subspace dimension by a number of eigenvalues, which include 99.9% signal power; thus this method can make a correct division between signal-subspace and noise-subspace. The other two methods. weighted noise-subspace method and extended MUSIC method, do not need to make a correct division between the two subspaces. Instead. These two methods can reduce the effect of signal-subspace dimension expanding directly by using weighted eigenvectors in the noise-subspace or the entire space. Computer simulations show that all these improved methods can overcome effectively the effect of signal-subspace dimension expanding of wideband signal-subspace spatial-spectrum estimation and possess good statistical performance.