Speaker Recognition Using Dynamic Features and The Null Space Combination of Classifiers in The Sub-band Domain