VoiceMind - AI for Healthcare

Problem Statement

Mental health disorders affect over 1 billion people globally, yet most go undetected because people cannot accurately self-report how they feel — traditional tools like mood questionnaires are subjective, inconsistent, and easily biased. Meanwhile, clinical research proves that vocal biomarkers (pitch, energy, speech rate, jitter) are objective, involuntary indicators of stress, anxiety, and depression — but this technology remains locked in expensive labs. VoiceMind bridges this gap: a free, AI-powered platform where users simply speak for 30 seconds, and our ML pipeline (librosa + Gemini AI) analyzes their voice patterns to detect emotional states, track wellness trends over time, and deliver personalized guidance — making mental health monitoring as effortless and routine as checking a step count, with zero self-reporting and zero barriers to access.

Solution & AI Usage

VoiceMind detects emotional states from a 30-second voice recording using real machine learning — no questionnaires, no self-reporting. A Python microservice (librosa) extracts vocal biomarkers: pitch, energy, speech rate, jitter, pause ratio, and MFCCs from the raw audio signal. These are objective, involuntary markers clinically linked to stress, anxiety, and depression. A scoring engine maps these features to emotional state classifications (0-100 scores). Google Gemini AI then interprets the raw metrics, generating personalized insights, risk flags, and actionable recommendations. The same AI powers a guardrailed wellness chat companion that suggests breathing exercises, grounding techniques, and habits — personalized to the user's detected emotional state. It never diagnoses and surfaces crisis resources when needed. The result: passive, objective mental health monitoring accessible to anyone with a microphone — completely free.

Full Description

Tech Stack