Analysis of fundamental frequency for near term suicidal risk assessment

Abstract

Among the many clinical decisions that psychiatrists must make, assessment of a patient's risk of committing suicide is definitely among the most important, complex and demanding. Clinical experience has shown that successful predictions of suicidality were often based on the patient's voice independent of content. Vocal patterns associated with dynamic expressiveness were designated as one of the most distinguishable features in the voices of suicidal patients at imminent risk. This paper investigates this phenomenon in an effort to explore the discriminating ability of jitter (period-to-period fluctuations in fundamental frequency) among suicidal and nonsuicidal patients. First, a wavelet transform based glottal cycle duration estimation technique is employed for the voiced/unvoiced decision and fundamental frequency (F0) estimation. Second, F0 perturbations are computed for each voiced segment in the F0 contour. Statistical analyses showed that F0 perturbations differ significantly (p=0.0069) between suicidal and nonsuicidal subjects' speech. Finally, a maximum likelihood classifier was developed, which obtained a correct classification score of 80%.