We have experience extracting and analyzing features from different biomedical data, and have recently emphasized voice analysis in our research:
Extraction and analysis of high-dimensional feature sets to characterize vocal production, speech patterns, and speech content is a promising direction for biomarker identification. Features characterizing vocal production are independent of speech content itself, and can provide objective measures of motor difficulties as well as objective means of assessing psychiatrically relevant states, such as mood and anxiety. Features related to speech patterns and content provide additional opportunities to characterize more complex emotional and cognitive states, as well as issues related to processing information and expressing thoughts. We will apply our analyses to voice data from thousands of children who have participated in a thorough battery of clinical assessments. By making these data, software, and results openly available, we will establish new normative standards for voice-based diagnosis, prediction of risk, and monitoring of symptom severity for a broad range of neuropsychiatric conditions. By relating these voice-derived features to ancillary data from brain imaging and other modalities, we will provide a richer context for the contribution of voice data to diagnosis and prediction.
As part of the Healthy Brain Network project, we are collecting structural and functional MR brain images, EEG, eye tracking data, and hours of interviews and mental health assessments from 10,000 children and adolescents (current n=1,200). We have recently added voice recording and will soon include actigraphy and physiological measures: Voice_recording: We collect audio recordings during all tests and interviews with a portable Sony ICD-UX 533 digital voice recorder. After participants watch a 4-minute emotional animated film called “The Present” in the MRI scanner, they narrate the story in their own words and answer a series of perspective-taking questions related to the film. During this narration and question answering session we collect high-definition video and high-fidelity audio recordings with a Røde NT1 cardioid condenser microphone. Actigraphy: Each participant will wear a wrist-worn ActiGraph wGT3X-BT to monitor movement throughout the day and night for up to one month. Physiological data: We are currently evaluating other wearable devices that help to infer internal state, such as electrodermal activity (stress) and photoplethysmography (heart rate) with Empatica devices. We will soon have participants wear the device while undertaking all assessments and interviews.