NeuroVibeNet

AI-Powered Multi-Modal Mental Health Screening & Wellness Monitoring Platform

Early Detection • Behavioral Analytics • Voice Analysis • Facial Emotion Recognition

An end-to-end AI platform that combines behavioral assessments, speech analysis, and facial emotion recognition to provide personalized mental wellness insights and early risk screening.

---

Overview

NeuroVibeNet is a full-stack AI-powered mental health screening platform designed to assist in the early identification of emotional distress, anxiety patterns, depressive tendencies, and behavioral changes through multi-modal data analysis.

The platform integrates:

• Behavioral and survey-based assessment
• Voice biomarker analysis
• Facial emotion recognition
• Longitudinal wellness tracking
• Personalized recommendations
• Clinical-style dashboard analytics

Rather than relying on a single data source, NeuroVibeNet combines multiple independent AI models and fuses their predictions using an ensemble decision engine to generate more reliable wellness insights.

Important: NeuroVibeNet is an educational and wellness-support platform. It is not intended to diagnose, treat, or replace professional medical advice.

---

Key Features

Multi-Step User Onboarding

• Secure account registration
• Demographic profiling
• Mental health history collection
• Emergency contact registration
• AI consent workflow

Behavioral Assessment Engine

• Dynamic mental health questionnaires
• Behavioral risk profiling
• Historical assessment tracking
• Automated feature extraction

Voice Analysis Module

• Browser-based audio recording
• MFCC feature extraction
• Acoustic biomarker analysis
• SVM + KNN hybrid classification

Facial Emotion Recognition

• Real-time video capture
• Facial landmark analysis
• Emotion recognition pipeline
• OpenCV-powered processing

Multi-Model Prediction Engine

Combines outputs from:

• Random Forest
• LightGBM
• Support Vector Machine (SVM)
• K-Nearest Neighbors (KNN)
• Facial Emotion Recognition Model

Using:

• Feature Engineering
• Recursive Feature Elimination
• Weighted Ensemble Voting

Wellness Dashboard

• Daily check-in tracking
• Mood journaling
• Activity streak calendar
• Progress visualization
• Personalized wellness recommendations

AI Generated Insights

Users receive:

• Risk assessment summaries
• Behavioral observations
• Wellness recommendations
• Daily action plans
• Do's and Don'ts suggestions

---

System Architecture

                    ┌──────────────────┐
                    │ React Frontend   │
                    └────────┬─────────┘
                             │
                             ▼
                    ┌──────────────────┐
                    │ Flask Backend    │
                    └────────┬─────────┘
                             │
          ┌──────────────────┼──────────────────┐
          ▼                  ▼                  ▼

   Behavioral Model     Audio Model      Video Model

(Random Forest +      (SVM + KNN)      Emotion Detection
    LightGBM)                            OpenCV Pipeline

          └──────────────────┬──────────────────┘
                             ▼

                    Fusion Engine
                  (Weighted Voting)

                             ▼

                 Final Risk Assessment

---

Machine Learning Pipeline

Behavioral Analysis

Input Sources:

• Survey responses
• Mood tracking
• Journal patterns
• User activity history

Techniques:

• Data preprocessing
• Feature encoding
• Scaling
• Missing value handling
• Recursive Feature Elimination (RFE)

Models:

• Random Forest
• LightGBM

---

Voice Biomarker Analysis

Audio Processing:

• Noise reduction
• Signal normalization
• Voice segmentation

Features Extracted:

• MFCCs
• Pitch
• Energy
• Spectral features
• Voice dynamics

Models:

• Support Vector Machine (SVM)
• K-Nearest Neighbors (KNN)

---

Facial Emotion Recognition

Video Processing:

• Frame extraction
• Facial landmark detection
• Emotion mapping
• Expression tracking

Technology:

• OpenCV
• ONNX Video Model

Outputs:

• Emotion probabilities
• Emotional trend analysis

---

Prediction Fusion

Predictions from all independent models are combined through:

• Confidence weighting
• Ensemble voting
• Risk score aggregation

This improves prediction stability and reduces reliance on a single modality.

---

Technology Stack

Frontend

Technology	Purpose
React.js	User Interface
Vite	Frontend Build Tool
Tailwind CSS	Styling
JavaScript	Application Logic
Recharts	Data Visualization
Chart.js	Analytics Dashboard

---

Backend

Technology	Purpose
Python	Core Backend
Flask	REST API Framework
MongoDB	Database
GridFS	Large Media Storage

---

Machine Learning

Technology	Purpose
Scikit-Learn	ML Models
LightGBM	Gradient Boosting
OpenCV	Computer Vision
ONNX Runtime	Model Inference
Librosa	Audio Processing
NumPy	Numerical Computing
Pandas	Data Processing

---

Project Structure

Neuro_Vibe_Net
│
├── frontend
│   ├── public
│   ├── src
│   │   ├── assets
│   │   ├── components
│   │   │   ├── AudioRecorder.jsx
│   │   │   ├── VideoRecorder.jsx
│   │   │   ├── MoodChart.jsx
│   │   │   ├── EmotionPieChart.jsx
│   │   │   ├── StreakCalendar.jsx
│   │   │   └── Recommendations.jsx
│   │   │
│   │   ├── context
│   │   ├── layouts
│   │   ├── pages
│   │   │   ├── Home.jsx
│   │   │   ├── Login.jsx
│   │   │   ├── Register.jsx
│   │   │   ├── Survey.jsx
│   │   │   ├── Dashboard.jsx
│   │   │   ├── RecordTest.jsx
│   │   │   ├── Results.jsx
│   │   │   └── AdminDashboard.jsx
│   │   │
│   │   └── services
│   │       └── api.js
│
├── backend
│   ├── database
│   ├── routes
│   ├── services
│   ├── models
│   ├── scripts
│   ├── uploads
│   │   ├── audio
│   │   └── video
│   │
│   ├── ml_models
│   │   ├── random_forest.pkl
│   │   ├── lightgbm_behavior.pkl
│   │   ├── svm_audio.pkl
│   │   ├── knn_audio.pkl
│   │   └── video_model.onnx
│   │
│   └── app.py
│
└── README.md

---

Security & Privacy

NeuroVibeNet was designed with privacy-first principles.

Security Measures

• Password-based authentication
• Encrypted API communication
• Secure media handling
• Isolated model inference pipeline
• Protected user assessment records

Privacy Measures

• Audio and video used only for analysis
• No public exposure of user data
• Controlled storage lifecycle
• Mental health data handled separately from model outputs

---

Platform Capabilities

1. Behavioral Assessment

2. Voice-Based Emotion Analysis

3. Facial Emotion Recognition

4. Multi-Model Ensemble Learning

5. Wellness Recommendations

6. Mood Journaling

7. Daily Check-ins

8. Streak Tracking

9. Admin Dashboard

10. MongoDB Media Storage

---

Installation

Clone Repository

git clone https://github.com/yourusername/Neuro_Vibe_Net.git
cd Neuro_Vibe_Net

---

Backend Setup

cd backend

python -m venv tfenv

# Windows
tfenv\Scripts\activate

# Linux / Mac
source tfenv/bin/activate

pip install -r requirements.txt

python app.py

Backend runs on:

http://localhost:5000

---

Frontend Setup

cd frontend

npm install

npm run dev

Frontend runs on:

http://localhost:5173

---

Application Screens

Landing Page

• Mental health platform introduction
• Call-to-action workflow

Registration Flow

• Multi-step onboarding
• Demographic collection
• Clinical history collection
• Consent management

Dashboard

• Wellness analytics
• Streak tracking
• Journal management
• Personalized recommendations

Voice & Video Assessment

• Real-time recording
• Multi-modal AI analysis

Results Engine

• AI-generated wellness insights
• Risk assessment output

---

Why This Project Matters

Mental health challenges often remain undetected until symptoms become severe.

NeuroVibeNet demonstrates how Artificial Intelligence, Machine Learning, Speech Processing, and Computer Vision can be combined into a practical healthcare-focused application capable of:

• Early risk identification
• Behavioral trend monitoring
• Wellness tracking
• Personalized intervention support

The project showcases full-stack development, machine learning engineering, MLOps concepts, data processing, API design, computer vision, audio analytics, and production-oriented system architecture.

---

Author

Vipul Paighan

AI Engineer • Full Stack Developer • Machine Learning Enthusiast

Focused on building intelligent systems that combine Artificial Intelligence with real-world impact.

---

License

This project is licensed under the MIT License.

---

Built with using React, Flask, Machine Learning, Computer Vision, and Audio Intelligence.