Working papers · speech, health & trustworthy AIVigo, 2026

Research / Mental-health AI and conversational safeguards / [8]

VisIA: detecting emotional distress and suicide risk from voice, face and text

VisIA Project: design of an automated AI-based emotional distress and suicide risk detection system. IberSPEECH 2024, 2024.

In one sentence

VisIA is a project to build a multimodal system that detects emotional distress in adolescents and a speech-based conversational agent that follows clinical guidelines, supported by a clinical trial planned for 339 participants in three groups.

339participants planned
3groups, 113 each
11–16age range

The problem

Suicide is one of the leading causes of death among young people, and current risk assessments produce both false positives and false negatives. VisIA is a joint project of clinicians and engineers that looks for a more accurate and non-invasive way to assess that risk.

What the project builds

Two systems. VisIA-Det is an automated detector of emotional distress that combines several kinds of data. VisIA-Bot is a speech-based conversational agent, designed according to clinical guidelines, to support people at risk. This paper concentrates on the detector.

The clinical trial

The trial is observational and prospective, with three groups of adolescents aged 11 to 16: those with current suicidal ideation who are receiving care, those with other psychological or psychiatric conditions and no suicidal ideation in the past year, and the general population. The planned sample is 339 participants, 113 per group. Clinical groups are recruited at two hospitals, in Ourense and Vigo, and the general population at five high schools.

The data

Participants complete validated screening questionnaires and answer a set of questions recorded as a video session and as written text. That gives three signals to analyse together: voice, facial expression and autobiographical text. A secure platform stores and integrates them.

Why it matters

The project’s questionnaire data are described in the VisIA-Q dataset.

Cite

Ramírez Sánchez J.M., Manso M., García-Mateo C., Gómez-Gómez B., Pinal B., Brañas A., García Caballero A., Docío-Fernández L., Fernández-Iglesias M.J. VisIA Project: design of an automated AI-based emotional distress and suicide risk detection system. IberSPEECH 2024, 2024.