An investigation on speaker diarization using support vector machines and autoassociative neural network