A Spectral Analysis of F0-contour in bipolar patients

Mental diseases are increasingly common. Among these, bipolar disorders heavily affect patients’ lives given the mood swings ranging from mania to depression. Voice has been shown to be an important cue to be investigated in this kind of diseases. In fact, several speech-related parameters were used to characterize voice in depressed people. The goal is to build a decision support system (DSS) improving diagnosis and possibly predicting mood changes. In the literature several works have been conducted regarding depression. Lately, some efforts were devoted to studies concerning bipolar patients. Here a spectral analysis of F0-contours extracted from audio recordings of text reading will be performed. The algorithm is completely automatic so that it can be easily integrated into a DSS. The proposed features are related to both speech rhythm and intonation. Analyses on both bipolar and healthy subjects are reported. The former ones were recorded while subjects were experiencing different mood states, while the latter were recorded at different days. Some coherent features trends are detected in bipolar patients across different mood states, while no significant differences are highlighted in healthy subjects. Preliminary results indicate that the proposed features could be fruitfully explored to characterize mood states in bipolar patients