Text-independent speaker identification based on piecewise canonical discriminant analysis | IEEE Conference Publication | IEEE Xplore