🧠 ManoVani – Conversational AI for Depression Detection

ManoVani is a voice-enabled AI system that intelligently detects signs of depression by analyzing both speech and text. It simulates an empathetic dialogue with users, offering support while classifying their mental state into:

• 😃 No Depression
• 🙂 Mild / False Depression
• 😔 True Depression

This project aims to assist early mental health diagnosis using machine learning and natural language processing, powered by the DAIC-WOZ dataset.

---

🌟 Features

• 🗣️ Voice Assistant (Sakhi)
• 📊 NBTree (Naive Bayes Tree) classifier
• 🎙️ Text + Speech input and output
• 🔐 Local prediction (Privacy-focused)
• 📁 Uses clinical DAIC-WOZ dataset
• 📈 PCA & Standard Scaler-based preprocessing

---

🧠 System Flow

graph TD;
    A[Voice/Text Input] --> B[Feature Extraction]
    B --> C[Scaling + PCA]
    C --> D[NBTree Model]
    D --> E[Depression Classification]
    E --> F[Voice Response Output]

📁Project Overview

ManoVani/
├── main.py                  # Core execution logic
├── sakhi.py                 # Voice interaction module
├── model.py                 # ML model loading and prediction
├── best_ensemble_model.pkl  # Trained NBTree model
├── scaler.pkl               # StandardScaler for normalization
├── pca.pkl                  # PCA model
├── processed_data.pkl       # Preprocessed features
├── response.mp3             # Final AI-generated response
├── requirements.txt         # Python dependencies
├── README.md                # GitHub project overview
├── index.md                 # GitHub Pages homepage (this file)
├── _config.yml              # GitHub Pages theme config
└── LICENSE                  # MIT License

🛠️ Tech Stack Python 3.11+

scikit-learn, NumPy, pandas

gTTS, SpeechRecognition, PyAudio

PCA, StandardScaler

DAIC-WOZ Dataset

🚀 Run Locally Clone the repository:

git clone https://github.com/NandiniJaiswal05/Fake-Depression-Detection-Using-Speech-Analysis.git
cd ManoVani

Install dependencies:

pip install -r requirements.txt

Start the application:

python main.py

📦 Dataset 🎯 Name: DAIC-WOZ

🎧 Audio + 📄 Text features

✅ Used: covarep.csv, formant.csv, transcript.csv, audio.wav

📌 Clinical dataset for mental health studies

📄 License This project is licensed under the MIT License.

⚠️ Disclaimer: ManoVani is a research-based tool and not a substitute for medical advice or treatment.

🙋‍♀️ Author Nandini Jaiswal AI/ML Developer | Mental Health Advocate 📍 GHRCE, India 🔗 GitHub

💡 Future Enhancements Multilingual Voice Support

Doctor/Counselor Integration

Emotional Progress Tracking

Streamlit/Web Frontend

Docker/Android Deployment

🔗 Live GitHub Pages Once deployed: 📍 https://NandiniJaiswal05.github.io/Fake-Depression-Detection-Using-Speech-Analysis