🎧 AudioCompass

AudioCompass is an open-source Arduino library for ESP32-S3 boards that detects the direction of sound using two I2S microphones, MFCC feature extraction, and a TinyML model. It tells you if a sound came from the Front, Back, Left, or Right.

---

🚀 Features

• ✅ Dual I2S microphone input (e.g., INMP441)
• ✅ Real-time audio buffering using I2S
• ✅ MFCC feature extraction using CMSIS-DSP
• ✅ Embedded TinyML model for sound direction
• ✅ Efficient TensorFlow Lite Micro inference
• ✅ Simple Arduino API: AudioCompass.getDirection()
• ✅ Fully Arduino Library Manager compatible

---

🔌 Hardware Requirements

Component	Quantity	Notes
ESP32-S3 Board	1	e.g., ESP32-S3 DevKit, etc.
I2S Mic	2	e.g., INMP441 or SPH0645
Jumper wires	—	To connect mics to I2S pins

---

📡 Wiring Example (ESP32-S3 + 2x INMP441)

Mic 1 (Left side):

INMP441 Pin	ESP32-S3 Pin
GND	GND
VCC	3.3V
WS	GPIO 15
SCK	GPIO 14
SD	GPIO 32

Mic 2 (Right side):

INMP441 Pin	ESP32-S3 Pin
GND	GND
VCC	3.3V
WS	GPIO 15 (shared)
SCK	GPIO 14 (shared)
SD	GPIO 33

✅ The WS (word select) and SCK (clock) lines are shared between both microphones.
SD (data line) must be unique for each mic.

---

📄 Example Sketch

#include <AudioCompass.h>

void setup() {
  Serial.begin(115200);
  AudioCompass.begin();
}

void loop() {
  String direction = AudioCompass.getDirection();
  Serial.println("Detected Direction: " + direction);
  delay(500);
}

---

🧠 Model Training Overview

The TinyML model was trained using real audio recorded from a stereo mic setup. Here's the process:

📊 1. Data Collection

Claps/beeps/taps were played from four angles (Front, Back, Left, Right)

Audio captured using two I2S mics ~20–40cm apart

Sampled at 16kHz

🧮 2. Feature Extraction

13 MFCCs per frame

30ms window, 10ms step

Normalized values

🤖 3. Model Architecture

python Copy Edit model = tf.keras.Sequential([ tf.keras.layers.Input(shape=(N_MFCC,)), tf.keras.layers.Dense(16, activation='relu'), tf.keras.layers.Dense(4, activation='softmax') ])

💾 4. Deployment

Trained model exported to .tflite

Converted to .h with xxd

Inference performed using TensorFlow Lite Micro

🧪 Inference Output

Calling AudioCompass.getDirection() will return a String:

"Front"

"Back"

"Left"

"Right"

These values are printed to Serial or can be used in your own logic.

📁 Folder Structure

File	Purpose
examples/AudioDirectionDemo.ino	Simple working example
src/model_data.h	Embedded TFLite model as a C array
src/AudioCompass.h/.cpp	Library implementation
library.properties	Arduino metadata
keywords.txt	Arduino IDE keyword coloring
README.md	This documentation
LICENSE	MIT License

---

📥 Installation

You can install this library via:

Arduino Library Manager (Recommended)

Or manually:

Download AudioCompass.zip

Arduino IDE > Sketch > Include Library > Add .ZIP Library

📜 License

sql Copy Edit MIT License

Copyright (c) 2025 Herobrine Pixel

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction...

[Full license text included in LICENSE file]

🙌 Credits

Developed by Herobrine Pixel with support from the OpenAI team. Includes CMSIS-DSP and TensorFlow Lite Micro runtime.

🔗 Useful Resources

ESP32 I2S Docs

CMSIS-DSP Library

TensorFlow Lite for Microcontrollers

Edge Impulse Studio