AI Security Alert System - Setup & Usage Guide

📋 Overview

This AI-powered security system monitors your camera and microphone to:

• Detect when anyone enters the camera view
• Recognize authorized persons (you)
• Alert when unauthorized persons are detected
• Monitor for loud sounds or unusual noises
• Provide real-time visual alerts

🔧 Installation

Step 1: Install Python Dependencies

# Install main requirements
pip install opencv-python numpy torch torchvision pyaudio pandas

# Or use the requirements file
pip install -r requirements.txt

Step 2: Platform-Specific Setup

Windows:

# PyAudio installation on Windows
pip install pipwin
pipwin install pyaudio

# Or download wheel from: https://www.lfd.uci.edu/~gohlke/pythonlibs/#pyaudio

macOS:

# Install PortAudio first
brew install portaudio

# Then install PyAudio
pip install pyaudio

Linux (Ubuntu/Debian):

# Install system dependencies
sudo apt-get update
sudo apt-get install python3-pyaudio portaudio19-dev

# Install Python packages
pip install pyaudio opencv-python

Step 3: Test Your Setup

# Test camera
python -c "import cv2; print('Camera:', cv2.VideoCapture(0).isOpened())"

# Test microphone
python -c "import pyaudio; p = pyaudio.PyAudio(); print('Microphone:', p.get_default_input_device_info())"

🚀 Usage

Basic Usage

python security_alert_system.py

First-Time Setup

1. Registration Phase:

   • Position yourself in front of the camera
   • Press SPACE to register your face as authorized
   • Press ESC to skip registration (system will alert for all persons)

2. Monitoring Phase:

   • The system starts monitoring automatically
   • Green box = Authorized person detected
   • Red box = Unauthorized person detected (ALERT!)
   • Sound monitoring runs in background

Keyboard Controls

Key	Action
q	Quit the system
r	Re-register authorized person
s	Adjust sound sensitivity

⚙️ Configuration

Adjust Sound Sensitivity

During runtime, press 's' and enter a value:

• 500-1500: Very sensitive (detects whispers)
• 1500-3000: Normal (detects conversations)
• 3000-5000: Less sensitive (only loud sounds)

Modify Alert Settings

Edit in security_alert_system.py:

# Line ~31: Alert cooldown (seconds between alerts)
self.alert_cooldown = 5  # Change to your preference

# Line ~26: Sound threshold
self.SOUND_THRESHOLD = 1000  # Adjust sensitivity

# Line ~19: Confidence threshold for person detection
self.person_detector.conf = 0.5  # 0.1 (sensitive) to 0.9 (strict)

📊 How It Works

1. Person Detection

• Uses YOLOv5 (You Only Look Once) for real-time person detection
• Processes each frame to identify human figures
• Draws bounding boxes around detected persons

2. Face Recognition

• Uses Haar Cascade Classifier for face detection
• Compares detected faces with registered authorized person
• Template matching to verify identity

3. Sound Monitoring

• Captures audio in real-time using PyAudio
• Calculates RMS (Root Mean Square) of audio signal
• Triggers alert when volume exceeds threshold

4. Alert System

• Visual alerts on screen (red boxes, text warnings)
• Console alerts with timestamp
• Cooldown period to prevent alert spam

🎯 Use Cases

Home Security

# Run continuously
python security_alert_system.py

Office Monitoring

# Detect when someone enters your workspace
# Register yourself as authorized person

Baby Monitor

# Detect movement and sounds in nursery
# Set lower sound threshold for sensitivity

🔒 Privacy & Security

• Local Processing: All AI processing happens on your device
• No Cloud Upload: Video/audio never sent to external servers
• No Recording: Doesn't save video unless you modify the code
• Authorized Person: Only you can register as authorized

🛠️ Customization Examples

Example 1: Save Alerts to File

Add to trigger_alert() method:

def trigger_alert(self, reason):
    # ... existing code ...
    
    # Save to log file
    with open('security_log.txt', 'a') as f:
        f.write(f"{timestamp} - {reason}\n")

Example 2: Save Snapshot on Alert

def trigger_alert(self, reason):
    # ... existing code ...
    
    # Save current frame
    cv2.imwrite(f'alert_{timestamp.replace(":", "-")}.jpg', self.frame_buffer[-1])

Example 3: Send Email Alert

import smtplib
from email.mime.text import MIMEText

def trigger_alert(self, reason):
    # ... existing code ...
    
    # Send email
    msg = MIMEText(f"Security Alert: {reason}")
    msg['Subject'] = 'Security Alert!'
    msg['From'] = 'your-email@gmail.com'
    msg['To'] = 'recipient@gmail.com'
    
    with smtplib.SMTP('smtp.gmail.com', 587) as server:
        server.starttls()
        server.login('your-email@gmail.com', 'your-password')
        server.send_message(msg)

Example 4: Multiple Authorized Persons

def register_authorized_person(self):
    # Modify to store multiple faces
    self.authorized_faces = []  # List of authorized faces
    
    # Allow registering multiple times
    # Store each face in the list

🐛 Troubleshooting

Camera Not Opening

# Check camera permissions
# Windows: Settings > Privacy > Camera
# macOS: System Preferences > Security & Privacy > Camera
# Linux: Check /dev/video0 permissions

PyAudio Installation Fails

# Windows: Use pipwin or download wheel
# macOS: Install portaudio via brew
# Linux: Install portaudio19-dev

YOLOv5 Download Slow

# It will auto-download on first run (~14MB)
# Be patient, it only happens once
# Or download manually from: https://github.com/ultralytics/yolov5

High CPU Usage

# Reduce frame processing in code:
self.cap.set(cv2.CAP_PROP_FRAME_WIDTH, 320)  # Lower resolution
self.cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 240)

# Or add frame skip:
if frame_count % 2 == 0:  # Process every 2nd frame
    # ... detection code ...

False Alarms

# Increase confidence threshold
self.person_detector.conf = 0.7  # Higher = stricter

# Adjust face matching threshold
if similarity > 0.7:  # Higher = stricter matching

📈 Performance Tips

1. Use USB Camera: Better quality than laptop webcam
2. Good Lighting: Improves person and face detection
3. Stable Position: Mount camera to reduce motion blur
4. Background: Plain backgrounds work better
5. Distance: Keep 3-6 feet from camera for best results

📝 System Requirements

• Python: 3.7 or higher
• RAM: Minimum 4GB (8GB recommended)
• CPU: Any modern processor (GPU optional)
• Camera: Any USB or built-in webcam
• Microphone: Optional but recommended
• OS: Windows 10/11, macOS 10.14+, Linux

🎓 Advanced Features (Optional)

Add SMS Alerts (Twilio)

pip install twilio

Add Database Logging (SQLite)

import sqlite3
# Store alerts in database

Add Web Dashboard (Flask)

pip install flask
# Create web interface to view alerts

Add Motion Detection

# Use frame differencing to detect movement
# Trigger alerts on significant motion

📞 Support

If you encounter issues:

1. Check the troubleshooting section above
2. Ensure all dependencies are installed
3. Verify camera and microphone permissions
4. Test each component individually

🔄 Updates & Improvements

To enhance the system:

• Replace Haar Cascade with deep learning face recognition
• Add object detection (not just persons)
• Implement motion tracking
• Add night vision support (IR camera)
• Create mobile app for remote monitoring

⚖️ Legal Notice

Important: Check local laws before using this system:

• Audio recording may require consent
• Video surveillance has different rules by location
• Always respect privacy rights
• Use responsibly and ethically

---

Happy Monitoring! 🔒