ESP32 Wake Word Detection

An ESP32-S3 based wake word detection system using Espressif's ESP-SR framework. Features voice control with animated face display feedback and supports both I2S digital and analog microphones.

Features

• Wake Word Detection: Activate with "Hi ESP" command
• Voice Commands: Control lights and fans with natural language
• Animated Display: OLED display with expressive face animations
• Dual Microphone Support: I2S digital or analog microphone options
• FreeRTOS Multi-Core: Optimized task distribution across both cores

Hardware Requirements

• ESP32-S3-DevKitC-1-N16R8 (16MB Flash, 8MB PSRAM)
• OLED Display (I2C, 128x64)
• Microphone Options:
  • I2S Digital: INMP441, ICS-43434, or SPH0645
  • Analog: MAX9814 or similar with gain control

Dependencies

• Platform: ESP32 Arduino (Custom platform with ESP-SR support)
• Libraries:
  • U8g2 (OLED Graphics)
  • ESP-SR (Speech Recognition)
  • ESP-Skainet (Wake Word Models)

Installation

1. Install PlatformIO (VSCode Extension or CLI)

2. Clone the repository:

   git clone https://github.com/jahrulnr/ESP32-WakeWord.git
   cd ESP32-WakeWord

3. Configure hardware settings in include/app_config.h:

   // Select microphone type
   #define MIC_TYPE MIC_TYPE_I2S  // or MIC_TYPE_ANALOG
   
   // Configure pins according to your wiring
   #define MIC_SCK GPIO_NUM_41    // I2S: SCK
   #define MIC_WS  GPIO_NUM_42    // I2S: WS
   #define MIC_DIN GPIO_NUM_2     // I2S: SD
   // Or for analog microphone
   #define MIC_OUT GPIO_NUM_4     // Analog: Output
   #define MIC_GAIN GPIO_NUM_38   // Analog: Gain

4. Build and upload:

   pio run -t upload

Voice Commands

1. Wake Word:

   • Say "Hi ESP" to activate command mode

2. Light Control:

   "Turn on the light" / "Switch on the light"
   "Turn off the light" / "Switch off the light" / "Go dark"
   

3. Fan Control:

   "Start fan"
   "Stop fan"
   

Project Structure

src/
├── main.cpp              # Entry point
├── boot/                 # System initialization
│   ├── setup.cpp        # Hardware setup sequence
│   └── constants.h      # Voice commands & constants
├── app/                 # Application logic
│   ├── tasks/          # FreeRTOS tasks
│   ├── callback/       # ESP-SR callbacks
│   └── display/        # Display functions
lib/                    # Custom libraries
├── Display/            # Display abstraction
├── FaceDisplay/        # Animated face system
├── Microphone/        # Microphone interfaces
└── Notification/      # Inter-task communication

⚡ Architecture

Task Distribution

• Core 0: Speech recognition processing

  • Priority 8
  • 4KB stack
  • Handles ESP-SR system and audio processing

• Core 1: Display and animations

  • Priority 19
  • 4KB stack
  • Manages UI updates and face animations

Memory Configuration

• Custom partition table (hiesp.csv)
• 8.9MB dedicated to model storage
• PSRAM optimization for ESP32-S3

Model Management

For detailed instructions on building, packaging, and flashing ESP-SR models, see the ESP-SR Model Management Guide. The guide includes:

• Model structure and organization
• Build and flash instructions
• Pre-built configurations
• Troubleshooting common model issues
• ESP-SR and I2S integration tips

Troubleshooting

1. Speech Recognition Issues

   • Verify 16kHz sample rate configuration
   • Check microphone signal levels
   • Monitor heap usage with serial output

2. Build Errors

   • Clean build: pio run -t clean
   • Verify ESP-SR platform compatibility
   • Check PSRAM configuration

3. Display Problems

   • Verify I2C pins in app_config.h
   • Check OLED address/configuration
   • Monitor display task stack usage

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• ESP-SR Framework by Espressif
• U8g2 Library
• Original face animations based on ESP32-Eyes