ESP32 Smart Presence Detection using WiFi RSSI and ML

Project Overview

This innovative project detects human presence in a room using WiFi signal strength (RSSI) and machine learning. It runs entirely on an ESP32 microcontroller with MicroPython, featuring an OLED display, buzzer, and SD card for real-time alerts and data logging - all without cameras for complete privacy.

Passive Detection

Uses existing WiFi signals to detect human presence without emitting any signals or requiring cameras.

Machine Learning

Implements trained models directly on the ESP32 to distinguish between empty and occupied rooms.

Real-time Alerts

Provides immediate visual (OLED) and audible (buzzer) alerts when presence is detected.

Data Logging

Records all detection events to SD card for later analysis in CSV format.

ESP32 Powered

Runs entirely on low-cost ESP32 hardware with MicroPython for easy development.

Privacy Focused

No cameras or personal data collection - only WiFi signal strength measurements.

Hardware Components

The system uses common, low-cost components that are easy to source and assemble:

Component	Description	Purpose
ESP32 Dev Board	WiFi-enabled microcontroller	Main processing unit with built-in WiFi
OLED Display	SSD1306 (128x64)	Real-time status display
Buzzer	PWM-controlled	Audible alerts for presence detection
Buttons	Tactile switches	User interface controls (select/back)
SD Card Module	SPI interface	Data logging storage
Power Supply	5V USB or battery	System power

Complete System Setup — Complete System Assembly

How It Works

The system operates through a carefully designed workflow that combines hardware sensing with machine learning:

Data Collection Phase

The ESP32 continuously scans for nearby WiFi networks and records their signal strength (RSSI) under two scenarios:

When the room is empty (blankroom.csv)
When a person is present in the room (personinroom.csv)

# Sample data collection code snippet
import network
import time
import json

wifi = network.WLAN(network.STA_IF)
wifi.active(True)

def scan_networks():
    networks = wifi.scan()
    rssi_values = [net[3] for net in networks]  # Extract RSSI values
    avg_rssi = sum(rssi_values) / len(rssi_values) if rssi_values else -100
    return {
        'timestamp_ms': time.ticks_ms(),
        'rssi': avg_rssi,
        'networks': len(networks)
    }

Machine Learning Workflow

Collect RSSI data under both scenarios
Train a machine learning model in Jupyter Notebook
Export a threshold-based or lightweight rule
Run real-time RSSI classification on ESP32
Trigger alert if human presence is detected

ML Model Training — Machine Learning Model Training

Data Analysis and Results

The project includes comprehensive Jupyter Notebook analysis to develop the optimal detection algorithm:

# blankroom.ipynb - Data Analysis

# Import libraries
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns

# Load and analyze data
df_blank = pd.read_csv("blankroom.csv")
df_person = pd.read_csv("personinroom.csv")

# Combine datasets
df_all = pd.concat([df_blank, df_person], ignore_index=True)

# Preprocessing
df_all["time_s"] = (df_all["timestamp_ms"] - df_all["timestamp_ms"].min()) / 1000
df_all["rssi_avg"] = df_all["rssi"].rolling(window=5, center=True).mean()
df_all = df_all.dropna()

# Plot RSSI distribution
plt.figure(figsize=(10,5))
sns.histplot(data=df_all, x="rssi_avg", hue="label", bins=30, kde=True)
plt.title("RSSI Distribution: Empty vs Person in Room")
plt.xlabel("RSSI (dBm)")
plt.ylabel("Frequency")
plt.show()

RSSI Over Time — RSSI Measurements Over Time

Model Performance

Accuracy: 0.9006

Classification Report:
              precision  recall  f1-score  support

           0       0.88      0.92      0.90       525
           1       0.92      0.88      0.90       521

    accuracy                           0.90      1046
   macro avg       0.90      0.90      0.90      1046
weighted avg       0.90      0.90      0.90      1046

Meet the Team

This project is developed by a passionate team dedicated to advancing IoT and machine learning technologies.

Mahendra Mali

Project Lead

IoT & ML Engineer

Barot JayRajKumar