⚠️ Notes

• Recommended Python version: 3.10
• PyAudio may not work on Python 3.12+

---

🎧 Infant Cry Analysis and Prediction System - ACADEMIC PROJECT

An AI-powered Final Year Engineering Project to analyze infant cry audio and predict the reason (hunger, discomfort, pain, tiredness) using Machine Learning.

---

📌 Problem Statement

Infants communicate through crying, but identifying the reason is difficult.
This project uses AI to classify cry sounds and assist caregivers.

---

🚀 Features

• 🎤 Audio input (microphone / dataset)
• 🔊 Noise reduction
• 📊 MFCC feature extraction
• 🤖 Random Forest classification
• 📈 Audio visualization
• 🌐 Flask-based web application (localhost)
• 🧠 Real-time prediction with probability distribution of cry reasons

---

🌐 Web Application (Flask)

A web-based interface is developed using Flask that runs on localhost.

Features:

• 🎙️ Record audio directly from microphone
• 📂 Upload audio file option
• ⚡ Real-time prediction
• 📊 Displays probability distribution of different cry types (hungry, pain, tired, discomfort)

Run the app:

python backend/app.py

Then open in browser:

http://127.0.0.1:5000/

---

🧠 Machine Learning Approach

• Feature Extraction: MFCC (Mel Frequency Cepstral Coefficients)
• Model: Random Forest Classifier
• Output: Predicted class + probability distribution
• Evaluation: Accuracy, Confusion Matrix

---

🏗️ Project Structure

Infant-Cry-Analysis-and-Prediction/ ├── audio_dataset/ ├── backend/ │ ├── app.py │ └── Cry_Model.ipynb ├── templates/ ├── requirements.txt └── README.md

---

⚙️ Setup Instructions

git clone https://github.com/YOUR_USERNAME/Infant-Cry-Analysis-and-Prediction.git
cd Infant-Cry-Analysis-and-Prediction

python -m venv venv
venv\Scripts\activate

pip install -r requirements.txt

python backend/app.py

---

📊 Tech Stack

Python, Flask, NumPy, Pandas, Librosa, Scikit-learn, Matplotlib, Seaborn

---

📊 Results

• 🎯 Model Accuracy: 90.47%
• ✅ Achieved high performance on test dataset
• 📈 Model performs well across most classes with balanced predictions

🔍 Performance Highlights:

• Strong precision and recall for:

  • hungry
  • burping
  • belly_pain

📉 Observations:

• Some confusion observed between:

  • discomfort ↔ tired

• Minor misclassifications due to similarity in cry patterns

📊 Sample Prediction Output:

hungry: 82%
tired: 10%
burping: 5%
discomfort: 2%
belly_pain: 1%

📌 Confusion Matrix Insights:

• High accuracy in identifying hungry and burping
• Slight performance drop in discomfort class

🎯 Final Outcome:

The model successfully predicts infant cry reasons with ~90% accuracy and provides probability-based outputs, making it useful for real-time caregiving assistance.