Evaluation of speaker normalization for suicidality assessment

When reviewing his clinical experience in treating suicidal patients, one of the authors observed that successful predictions of suicidality were often based on the patient’s voice independent of content. Using the Gaussian mixture model to represent the mel-cepstral features of voiced speech, speech of suicidal persons can be distinguish from that of depressed and control persons. The question then becomes can warping of the frequency axis improve the classification. The results show that warping of the frequency axis using the third format or Gaussian mixture model technique produces the best classification results.