Voice acoustical measurement of the severity of major depression

A number of empirical studies have documented the relationship between quantifiable and objective acoustical measures of voice and speech, and clinical subjective ratings of severity of Major Depression. To further explore this relationship, speech samples were extracted from videotape recordings of structured interviews made during the administration of the 17-item Hamilton Depression Rating Scale (HDRS; Hamilton, 1960). Pilot data were obtained from seven subjects (five males, two females) from videotapes that have been used to train expert raters on the administration and scoring of the HDRS. Several speech samples were isolated for each subject and processed to obtain the acoustic measurements. Acoustic measures were selected on the basis that they were correlated with HDRS ratings of symptom severity as seen under ideal voice recording conditions in previous studies. Our findings corroborate earlier reports that speaking rate is well correlated (negatively) with HDRS scores, with a strong correlation and nearly significant trend seen for the measure of pitch variability. A moderate pairwise correlation between percent pause time and HDRS score was also revealed, although this relationship was not statistically significant. The results from this cross-sectional study further demonstrate the ability of voice and speech signal analyses to objectively track severity of depression. In the present case, it is suggested that this relationship is robust enough to be found despite the less than ideal recording conditions and equipment used during the original videotape recording. Voice acoustical analyses may provide a powerful compliment to the standard clinical interview for depression. Use of such measures increases the range of techniques that are available to explore the neurobiological substrates of Major Depression, its treatment, and the dynamic interplay of the systems that govern the motor, cognitive, and emotional aspects of speech production. (C) 2004 Elsevier Inc. All rights reserved.