Emotion Detection Project

This project implements a real-time emotion detection system using AI models and MQTT communication. It includes:

• emotion_demo.py: Core AI script for emotion detection.
• printResult.py: Script to display results and handle control commands via MQTT.

---

Table of Contents

1. Prerequisites
2. Usage
   • Start Mosquitto Broker
   • Run the Emotion Detection Script
   • Run the Result Display Script
3. File Details
4. Troubleshooting
5. License

---

Prerequisites

1. Python Installation Ensure Python (>=3.7) is installed. Download Python.

2. Mosquitto Installation Install Mosquitto MQTT broker version 2.0.12.

   Note: The latest versions may not support WebSocket communication. Version 2.0.12 is confirmed to work.

   • Windows: Download Mosquitto 2.0.12.
   • Linux:

     sudo apt update
     sudo apt install -y mosquitto=2.0.12 mosquitto-clients

   • MacOS:

     brew install mosquitto@2.0.12

   To verify the installed version:

   mosquitto -v

3. Dependencies Installation Install the required Python libraries:

   pip install -r requirements.txt

---

Usage

1. Start Mosquitto Broker

Run the Mosquitto MQTT broker with WebSocket support enabled (default port 8083):

mosquitto

2. Run the Emotion Detection Script

Execute the emotion_demo.py script to start the emotion detection:

python emotion_demo.py

3. Run the Result Display Script

Run printResult.py to subscribe to the MQTT topic and display results:

python printResult.py

---

File Details

• emotion_demo.py:

  • Loads a pre-trained model (emotion_detection.h5) and uses a Haar Cascade classifier (haarcascade_frontalface_default.xml) for face detection.
  • Publishes results via MQTT.
  • Update the absolute path to the model if necessary:

    classifier = load_model('C:/Users/manhh/Downloads/emotion detect/emotion_detection.h5')

• printResult.py:

  • Subscribes to two MQTT brokers:
    • Local MQTT broker (TCP, port 1883).
    • WebSocket MQTT broker (port 8083).
  • Dynamically switches between local MQTT and WebSocket brokers based on control commands.
    • Control Commands: Received via WebSocket, with key ai to toggle AI mode:
      • ai: true: Switches to local MQTT (1883) for emotion messages.
      • ai: false: Switches back to WebSocket (8083).
  • Forwards messages from WebSocket broker to local MQTT when in WebSocket mode.
  • Example processing flow:
    1. Receives JSON messages containing emotion data or control commands.
    2. Decodes and processes messages based on their source.
    3. Prints processed emotions or logs forwarded messages.

• emotion_detection.h5: Pre-trained model for emotion classification.

• haarcascade_frontalface_default.xml: Haar Cascade XML for face detection.

• requirements.txt: Python dependencies for the project.

---

Troubleshooting

1. Mosquitto not running: Ensure Mosquitto version 2.0.12 is installed and running:

   mosquitto

2. WebSocket issues: Verify that Mosquitto is configured to listen on port 8083. Check the configuration file (usually located at /etc/mosquitto/mosquitto.conf) and ensure it includes:

   listener 8083
   protocol websockets
   

3. Missing dependencies: Reinstall the dependencies:

   pip install -r requirements.txt

4. FileNotFoundError: Verify all required files (emotion_detection.h5, haarcascade_frontalface_default.xml) are in the correct directory.

---

Example Output

From printResult.py:

[MQTT LOCAL] Message received on emotion_topic: {"emotion": "Happy"}
[MQTT LOCAL] Processed Emotion: Happy
[MQTT WS] Forwarded message to MQTT local: {"emotion": "Neutral"}
[MQTT WS] Switching to MQTT (1883) mode (ai = true).

---

License

Feel free to use and modify this file as needed.