SPEAKER IDENTIFICATION USING SELECTIVE COMPARISON OF PITCH CONTOUR PARAMETERS

A method of selective pitch data comparison for speaker identification is presented. Pitch parameters of rising and falling nuclear monosyllables and filled hesitation pauses are evaluated for their discriminating ability using Fisher’s ratio and equal error rate measures obtained on a 10-male 3-session database of neutral-mode spontaneous statements in Tajik. “Physical ” pitch parameters, the best of which provided 75-78 % speaker discrimination accuracy in isolation, proved more effective than linguistically conditioned ones. Using all pitch parameters in combination produced identification accuracy of 87%. The ultimate aim of the present work is to develop an algorithm for automated expert assessment of overall pitch pattern similarity in speech samples for its further application within a multifold speaker identification system