Below is a complete reference design you can use as a final-year/academic IoT project, prototype, GitHub README, or implementation document. It combines ESP32, microphone-based baby-cry detection, sleep/activity analysis, n8n automation, an AI agent, Telegram voice alerts, Google Sheets, ThingSpeak, and a web dashboard.
AI Smart Cradle with Baby Cry Detection & Sleep Analysis
1. Project Overview
The AI Smart Cradle is an IoT-enabled intelligent baby-monitoring and automated cradle system designed to detect baby crying, monitor environmental conditions and cradle movement, estimate sleep/activity states, and automatically notify a parent or caregiver.
The system uses an ESP32 as the edge controller. Sensors collect information such as sound level, temperature, humidity, motion and cradle movement. A lightweight AI/ML model can classify audio into categories such as:
-
Baby crying
-
Baby vocalization
-
Normal/background sound
-
Other/noise
The ESP32 sends important events to an n8n automation server. n8n acts as the central workflow/orchestration layer and can:
-
Receive sensor data from ESP32.
-
Determine whether an event requires attention.
-
Store measurements in Google Sheets.
-
Send data to ThingSpeak.
-
Invoke an AI agent for event interpretation.
-
Send Telegram notifications.
-
Generate or deliver a voice alert.
-
Maintain an event history.
-
Control the cradle according to predefined safety rules.
The system is designed as a monitoring and assistance system, not as a replacement for parental supervision or medical monitoring.
2. Main Objectives
The project has the following objectives:
Hardware objectives
-
Monitor baby crying.
-
Monitor temperature and humidity.
-
Detect cradle movement.
-
Detect whether the baby/cradle is moving.
-
Provide local indication using LEDs/buzzer/display.
-
Provide optional automatic rocking.
AI objectives
-
Detect crying from microphone/audio features.
-
Distinguish crying from ordinary environmental noise.
-
Estimate sleep/activity state using sensor history.
-
Detect repeated crying events.
-
Provide an AI-generated event summary.
IoT objectives
-
Connect ESP32 to Wi-Fi.
-
Send telemetry to the cloud.
-
Maintain historical data.
-
Provide a web dashboard.
-
Integrate ThingSpeak.
Automation objectives
-
Use n8n as the workflow engine.
-
Trigger notifications automatically.
-
Store events in Google Sheets.
-
Send Telegram alerts.
-
Generate voice notifications.
-
Provide an AI-agent interface.
3. High-Level Architecture
┌─────────────────────────┐
│ BABY / CRADLE │
└────────────┬────────────┘
│
┌──────────────────┼──────────────────┐
│ │ │
▼ ▼ ▼
Microphone Temperature Motion/
/ Audio Sensor & Humidity Vibration
│ │ │
└──────────────────┼──────────────────┘
▼
┌─────────────────────────┐
│ ESP32 │
│ │
│ Sensor acquisition │
│ Edge processing │
│ Cry detection │
│ Sleep/activity logic │
│ Wi-Fi communication │
└────────────┬────────────┘
│
Wi-Fi / HTTP / MQTT
│
▼
┌─────────────────────────┐
│ n8n │
│ Automation Server │
│ │
│ Webhook │
│ Rules │
│ AI Agent │
│ Notifications │
│ Data processing │
└──────┬───────┬──────────┘
│ │
┌─────────────┘ └──────────────┐
▼ ▼
┌──────────────┐ ┌──────────────┐
│ Google Sheets│ │ ThingSpeak │
│ Event Log │ │ IoT Charts │
└──────────────┘ └──────────────┘
│
▼
┌───────────────┐
│ AI Agent │
│ Event analysis│
└───────┬───────┘
│
▼
┌───────────────┐
│ Telegram │
│ Text + Voice │
│ Notifications │
└───────────────┘
4. Complete System Flow
START
│
▼
ESP32 boots
│
▼
Connect to Wi-Fi
│
▼
Initialize sensors
│
▼
Read microphone/audio
│
├───────────────► Extract audio features
│ │
│ ▼
│ Cry classifier
│ │
│ ▼
│ Cry probability
│
├───────────────► Temperature/Humidity
│
├───────────────► Motion/Vibration
│
▼
Combine sensor information
│
▼
Calculate current state
│
├── Normal
├── Sleeping
├── Moving
├── Possible crying
└── Repeated crying
│
▼
Send JSON to n8n
│
▼
n8n Webhook
│
▼
Validate data
│
├──► Google Sheets
│
├──► ThingSpeak
│
└──► AI Agent
│
▼
Analyze event
│
▼
Is alert required?
/ \
NO YES
│ │
▼ ▼
Log only Telegram message
│
▼
Voice alert
│
▼
Parent
5. Hardware Components
Required components
| Component | Purpose |
|---|---|
| ESP32 DevKit | Main controller |
| Microphone/audio sensor | Baby-cry detection |
| DHT22/SHT31 | Temperature/humidity |
| MPU6050 | Motion/tilt detection |
| Vibration sensor | Cradle vibration/activity |
| OLED display | Local status |
| LED | Status indication |
| Buzzer | Local warning |
| 5V power supply | System power |
| Optional motor | Automatic rocking |
| Motor driver | Motor control |
| Wi-Fi router | Internet connectivity |
For an improved audio system, an I2S MEMS microphone is preferable to a basic analog sound sensor because it provides actual audio samples suitable for feature extraction and ML classification.
6. Suggested ESP32 Hardware Architecture
ESP32
┌────────────────┐
│ │
I2S Mic ────►│ I2S │
│ │
DHT22 ──────►│ GPIO │
│ │
MPU6050 ────►│ I2C │
│ │
OLED ───────►│ I2C │
│ │
Vibration ──►│ GPIO │
│ │
LED ◄────────│ GPIO │
│ │
Buzzer ◄─────│ GPIO │
│ │
Motor Driver◄│ PWM/GPIO │
│ │
Wi-Fi ◄─────►│ Wi-Fi │
└────────────────┘
7. Example Pin Configuration
A possible ESP32 configuration is:
| Device | ESP32 pin |
|---|---|
| DHT22 DATA | GPIO 4 |
| MPU6050 SDA | GPIO 21 |
| MPU6050 SCL | GPIO 22 |
| Vibration sensor | GPIO 27 |
| Status LED | GPIO 2 |
| Buzzer | GPIO 26 |
| Motor driver IN1 | GPIO 18 |
| Motor driver IN2 | GPIO 19 |
| OLED SDA | GPIO 21 |
| OLED SCL | GPIO 22 |
For an I2S microphone, choose the I2S pins according to the microphone module and firmware configuration.
Important: Do not connect a cradle motor directly to an ESP32 GPIO. Use an appropriate motor driver, separate motor power supply, flyback protection where applicable, and suitable mechanical safety limits.
8. Electrical Block Diagram
5V POWER SUPPLY
│
┌──────────┴──────────┐
│ │
▼ ▼
ESP32 Motor Driver
│ │
┌─────┼──────┐ ▼
│ │ │ Motor
│ │ │
▼ ▼ ▼
DHT22 MPU6050 Microphone
│ │ │
│ │ └──── Audio
│ │
│ └──── Motion
│
└──────── Temperature
9. Baby Cry Detection
The microphone continuously samples the surrounding sound.
A basic system can calculate:
-
RMS energy
-
Zero-crossing rate
-
Spectral centroid
-
Spectral bandwidth
-
Mel-frequency features
-
MFCC features
A machine-learning model can then classify an audio window.
Example:
Audio
│
▼
Sampling
│
▼
Noise filtering
│
▼
Audio window
│
▼
Feature extraction
│
├── RMS
├── ZCR
├── MFCC
└── Spectral features
│
▼
ML classifier
│
▼
Probability
│
├── Cry = 0.91
├── Normal = 0.06
└── Noise = 0.03
│
▼
Cry detected
A threshold such as 0.80 can be used in a prototype, but it should be determined experimentally from a properly collected and labelled dataset rather than assumed to be universally accurate.
10. AI Model
A practical architecture is:
Microphone
│
▼
Audio preprocessing
│
▼
MFCC / Mel spectrogram
│
▼
Small neural network
│
▼
Classification
│
┌───┼─────────────┐
▼ ▼ ▼
Cry Vocalization Noise
Possible model approaches include:
-
TensorFlow Lite Micro
-
Edge Impulse
-
Custom small neural network
-
Logistic regression/SVM for a simpler prototype
For an ESP32, the model should be kept small enough for available RAM/flash and inference time.
11. Sleep Analysis
The system should describe sleep as an estimated state, not medically diagnose sleep.
For example:
Sensor history
│
├── Sound level
├── Cry events
├── Movement
├── Cradle activity
└── Time
│
▼
Feature aggregation
│
▼
Sleep-state algorithm
│
┌─────┼─────────────┐
▼ ▼ ▼
Quiet Light activity Active
sleep
Example state logic:
If movement is low
AND cry probability is low
AND sound level is low
for a continuous interval:
estimated_state = "QUIET_SLEEP"
If movement increases:
estimated_state = "ACTIVE"
If repeated high cry probability occurs:
estimated_state = "CRYING"
The algorithm should use time windows rather than making a sleep-state decision from one sensor reading.
12. Example JSON Sent from ESP32
{
"device_id": "SMART_CRADLE_01",
"timestamp": 1750000000,
"temperature": 27.4,
"humidity": 61.2,
"motion": 0.18,
"sound_level": 54,
"cry_probability": 0.91,
"cry_detected": true,
"sleep_state": "CRYING",
"battery": 87
}
This JSON becomes the primary data structure passed to n8n.
13. n8n Architecture
n8n is the central automation layer.
ESP32
│
│ HTTP POST
▼
Webhook
│
▼
Validate JSON
│
▼
Normalize data
│
├───────────────┐
│ │
▼ ▼
Google Sheets ThingSpeak
│
▼
Decision/Routing
│
▼
AI Agent
│
▼
Alert decision
│
▼
Telegram
│
▼
Voice notification
14. n8n Workflow
Recommended nodes:
-
Webhook
-
Code / Set
-
IF
-
Google Sheets
-
HTTP Request – ThingSpeak
-
AI Agent
-
Telegram
-
HTTP Request/TTS service
-
Telegram voice/audio
-
Response
Example:
[Webhook]
|
v
[Validate JSON]
|
v
[Store Event]
|
+------> [Google Sheets]
|
+------> [ThingSpeak]
|
v
[IF Cry Probability > Threshold]
|
YES
|
v
[AI Agent]
|
v
[Generate Alert]
|
v
[Telegram Text]
|
v
[Text-to-Speech]
|
v
[Telegram Voice]
15. AI Agent
The AI agent should not directly make unrestricted hardware decisions.
A safer architecture is:
ESP32
│
▼
n8n validation
│
▼
Rule engine
│
▼
AI Agent
│
▼
Structured response
│
▼
Allowed action list
Example AI input:
Temperature: 27.4 C
Humidity: 61%
Cry probability: 0.91
Motion: low
Cry duration: 18 seconds
Previous cry events: 3
Example structured response:
{
"event": "possible_baby_cry",
"priority": "high",
"message": "Possible baby crying detected for approximately 18 seconds.",
"recommended_action": "check_baby"
}
The AI agent should be prevented from issuing unsafe commands such as uncontrolled motor operation.
16. Telegram Alert
Example text alert:
🚨 Smart Cradle Alert
Possible baby crying detected.
Cry probability: 91%
Temperature: 27.4°C
Humidity: 61%
Movement: Low
Estimated state: Crying
Please check the baby.
The voice notification can communicate the same information in a concise form.
Example:
"Smart Cradle alert. Possible baby crying has been detected. Please check the baby."
17. Telegram Conversation
A useful Telegram interface can support commands such as:
Parent:
/status
Smart Cradle:
Temperature: 27.4°C
Humidity: 61%
State: Quiet Sleep
Cry probability: 4%
Last event: 12 minutes ago
Another example:
Parent:
/today
Smart Cradle:
Today's summary:
Cry events: 7
Estimated quiet periods: 5
Movement events: 14
Average temperature: 27.1°C
Average humidity: 60.4%
AI-agent interaction:
Parent:
Why did I receive the last alert?
AI Agent:
The system detected a high cry probability for approximately
18 seconds and generated an alert because the event exceeded
the configured threshold.
18. Google Sheets Database
Create columns such as:
| Timestamp | Device | Temperature | Humidity | Motion | Cry Probability | Cry | Sleep State | Alert |
|---|---|---|---|---|---|---|---|---|
| 21:10 | CRADLE01 | 27.2 | 60 | 0.1 | 0.03 | No | Sleep | No |
| 21:20 | CRADLE01 | 27.3 | 61 | 0.2 | 0.91 | Yes | Crying | Yes |
Google Sheets is useful for:
-
Prototype logging
-
Project demonstrations
-
Data analysis
-
Generating graphs
-
Training-data collection
-
Event history
For a production-scale system, a proper time-series/database backend would generally be more appropriate.
19. ThingSpeak
ThingSpeak can be used for time-series visualization.
Possible fields:
Field 1 = Temperature
Field 2 = Humidity
Field 3 = Motion
Field 4 = Sound Level
Field 5 = Cry Probability
Field 6 = Sleep State
Example :
ESP32
│
▼
n8n
│
▼
HTTP Request
│
▼
ThingSpeak
│
▼
Charts
20. IoT Web Dashboard
The webpage can contain:
┌───────────────────────────────────────────────┐
│ AI SMART CRADLE │
├───────────────────────────────────────────────┤
│ Temperature Humidity Baby State │
│ 27.4 °C 61 % SLEEPING │
├───────────────────────────────────────────────┤
│ Cry Probability │
│ ███████████░░░░░░░░ 42% │
├───────────────────────────────────────────────┤
│ LIVE SENSOR DATA │
│ Temperature ──────────────── │
│ Humidity ──────────────── │
│ Cry level ──────────────── │
│ Movement ──────────────── │
├───────────────────────────────────────────────┤
│ Recent Events │
│ 21:15 Cry detected │
│ 21:08 Quiet sleep │
│ 20:55 Movement detected │
└───────────────────────────────────────────────┘
The webpage can obtain data through:
ESP32 → n8n → Database/API → Webpage
or:
ESP32 → ThingSpeak → Webpage
21. ESP32 Firmware
The ESP32 firmware has five major responsibilities:
setup()
│
├── Wi-Fi
├── sensors
├── display
└── audio
│
▼
loop()
│
├── Read sensors
├── Process audio
├── Determine state
├── Send telemetry
└── Wait
A minimal ESP32 telemetry prototype can be implemented as follows.
#include <WiFi.h>
#include <HTTPClient.h>
#include <DHT.h>
#include <Wire.h>
#define DHT_PIN 4
#define DHT_TYPE DHT22
#define MOTION_PIN 27
#define LED_PIN 2
const char* WIFI_SSID = "YOUR_WIFI";
const char* WIFI_PASSWORD = "YOUR_PASSWORD";
const char* N8N_WEBHOOK =
"https://YOUR_N8N_SERVER/webhook/smart-cradle";
DHT dht(DHT_PIN, DHT_TYPE);
unsigned long lastSend = 0;
const unsigned long SEND_INTERVAL = 10000;
float estimateCryProbability()
{
/*
Replace this function with the actual ML model.
This placeholder demonstrates the software architecture.
*/
int motion = digitalRead(MOTION_PIN);
if (motion == HIGH)
return 0.05;
return 0.02;
}
String determineSleepState(float cryProbability, int motion)
{
if (cryProbability > 0.80)
return "CRYING";
if (motion == HIGH)
return "ACTIVE";
return "QUIET_SLEEP";
}
void sendTelemetry()
{
float temperature = dht.readTemperature();
float humidity = dht.readHumidity();
if (isnan(temperature) || isnan(humidity))
return;
int motion = digitalRead(MOTION_PIN);
float cryProbability = estimateCryProbability();
bool cryDetected = cryProbability >= 0.80;
String sleepState =
determineSleepState(cryProbability, motion);
String json = "{";
json += "\"device_id\":\"SMART_CRADLE_01\",";
json += "\"temperature\":" + String(temperature, 2) + ",";
json += "\"humidity\":" + String(humidity, 2) + ",";
json += "\"motion\":" + String(motion) + ",";
json += "\"cry_probability\":" +
String(cryProbability, 3) + ",";
json += "\"cry_detected\":" +
String(cryDetected ? "true" : "false") + ",";
json += "\"sleep_state\":\"" + sleepState + "\"";
json += "}";
HTTPClient http;
http.begin(N8N_WEBHOOK);
http.addHeader("Content-Type", "application/json");
int responseCode = http.POST(json);
Serial.print("HTTP response: ");
Serial.println(responseCode);
http.end();
}
void setup()
{
Serial.begin(115200);
pinMode(MOTION_PIN, INPUT);
pinMode(LED_PIN, OUTPUT);
dht.begin();
WiFi.begin(WIFI_SSID, WIFI_PASSWORD);
Serial.print("Connecting to Wi-Fi");
while (WiFi.status() != WL_CONNECTED)
{
delay(500);
Serial.print(".");
}
Serial.println();
Serial.println("Wi-Fi connected");
Serial.println(WiFi.localIP());
}
void loop()
{
if (millis() - lastSend >= SEND_INTERVAL)
{
lastSend = millis();
sendTelemetry();
}
}
This is the IoT communication prototype. The estimateCryProbability() function must be replaced by an actual trained audio classifier for genuine AI-based cry detection.
22. n8n Webhook Input
The webhook receives:
{
"device_id": "SMART_CRADLE_01",
"temperature": 27.4,
"humidity": 61.2,
"motion": 0,
"cry_probability": 0.91,
"cry_detected": true,
"sleep_state": "CRYING"
}
A Code node can normalize the information:
const data = $json;
const cryProbability =
Number(data.cry_probability || 0);
const alertRequired =
data.cry_detected === true &&
cryProbability >= 0.80;
return [{
json: {
...data,
alert_required: alertRequired,
received_at: new Date().toISOString()
}
}];
23. n8n Decision Logic
cry_probability >= 0.80?
│
┌───┴───┐
│ │
YES NO
│ │
▼ ▼
Check Log only
duration
│
▼
Repeated event?
│ │
YES NO
│ │
▼ ▼
Alert Log
A duration requirement is useful because it can reduce false alerts caused by a short sound.
24. Example n8n AI-Agent Prompt
You are the event-analysis component of an AI smart cradle.
Analyze the supplied sensor event.
You must:
1. Describe the event using only the supplied data.
2. Never claim that the system has medically diagnosed the baby.
3. Never claim certainty when the classifier only provides a probability.
4. Recommend that a caregiver check the baby when a sustained high-probability
crying event is detected.
5. Return concise structured JSON.
Input:
Temperature: {{$json.temperature}}
Humidity: {{$json.humidity}}
Motion: {{$json.motion}}
Cry probability: {{$json.cry_probability}}
Cry detected: {{$json.cry_detected}}
Sleep state: {{$json.sleep_state}}
Return:
{
"event": "...",
"priority": "...",
"message": "...",
"recommended_action": "..."
}
25. Telegram Bot Integration
The Telegram workflow can be:
n8n
│
▼
Telegram node
│
├── Text message
│
└── Voice/audio message
Example alert:
🚨 AI SMART CRADLE
Possible crying detected.
Cry probability: 91%
Duration: 18 seconds
Temperature: 27.4°C
Humidity: 61%
Please check the baby.
26. Voice Alert Pipeline
AI Agent
│
▼
Alert text
│
▼
Text-to-Speech
│
▼
Audio file
│
▼
Telegram
│
▼
Parent's phone
The TTS provider can be selected according to project requirements. The n8n workflow only needs an HTTP/API step between the generated text and Telegram.
27. Web Dashboard API
A simple backend endpoint can return:
{
"temperature": 27.4,
"humidity": 61.2,
"cry_probability": 0.04,
"sleep_state": "QUIET_SLEEP",
"last_alert": "2026-09-28T20:52:00",
"device_status": "ONLINE"
}
The webpage can periodically request this endpoint.
Example JavaScript:
async function updateDashboard() {
const response = await fetch("/api/status");
const data = await response.json();
document.getElementById("temperature").textContent =
data.temperature + " °C";
document.getElementById("humidity").textContent =
data.humidity + " %";
document.getElementById("state").textContent =
data.sleep_state;
document.getElementById("cry").textContent =
Math.round(data.cry_probability * 100) + "%";
}
setInterval(updateDashboard, 5000);
updateDashboard();
28. Example HTML Dashboard
<!DOCTYPE html>
<html>
<head>
<title>AI Smart Cradle</title>
<style>
body {
font-family: Arial, sans-serif;
background: #eef4ff;
margin: 0;
padding: 30px;
}
h1 {
color: #183153;
}
.grid {
display: grid;
grid-template-columns:
repeat(auto-fit, minmax(200px, 1fr));
gap: 20px;
}
.card {
background: white;
padding: 25px;
border-radius: 15px;
box-shadow: 0 4px 15px #0002;
}
.value {
font-size: 32px;
font-weight: bold;
color: #2463eb;
}
</style>
</head>
<body>
<h1>AI Smart Cradle</h1>
<div class="grid">
<div class="card">
<h3>Temperature</h3>
<div class="value" id="temperature">--</div>
</div>
<div class="card">
<h3>Humidity</h3>
<div class="value" id="humidity">--</div>
</div>
<div class="card">
<h3>Cry Probability</h3>
<div class="value" id="cry">--</div>
</div>
<div class="card">
<h3>Baby State</h3>
<div class="value" id="state">--</div>
</div>
</div>
<script src="dashboard.js"></script>
</body>
</html>
29. Complete Data Flow
┌──────────────┐
│ Microphone │
└──────┬───────┘
│
▼
┌──────────────┐
│ ESP32 │
│ Edge AI │
└──────┬───────┘
│
JSON over Wi-Fi
│
▼
┌──────────────┐
│ n8n │
└──────┬───────┘
│
┌─────────────┼──────────────┐
│ │ │
▼ ▼ ▼
Google ThingSpeak AI Agent
Sheets │ │
│ ▼
│ Alert decision
│ │
└──────┬───────┘
▼
Telegram
/ \
Text Voice
│ │
└────┬────┘
▼
Parent
30. Project Operating Modes
Normal mode
No crying
↓
Log sensor values
↓
Update dashboard
↓
No alert
Cry detection mode
Cry probability > threshold
↓
Start event timer
↓
Confirm persistence
↓
Generate alert
↓
Telegram
↓
Voice notification
Quiet sleep mode
Low sound
+
Low movement
+
No detected crying
+
Sufficient observation time
↓
Estimated quiet sleep
Repeated crying mode
Cry event
↓
Recovery
↓
Cry event
↓
Recovery
↓
Cry event
↓
Repeated-event rule
↓
Higher-priority notification
31. False-Positive Reduction
This is an important part of the project.
Do not trigger an alert from one noisy audio sample.
Instead:
Audio detected
│
▼
Classifier
│
▼
Probability > threshold?
│
YES
│
▼
Persistence timer
│
▼
Still crying?
│
YES
│
▼
Generate alert
Potential sources of false positives include:
-
Television
-
Music
-
Adult speech
-
Dogs
-
Door sounds
-
Fan noise
-
Motor noise
-
Other babies
-
Sudden environmental sounds
Training data should include these negative examples.
32. AI Training Dataset
Create a labelled dataset with categories such as:
dataset/
│
├── crying/
│ ├── cry001.wav
│ ├── cry002.wav
│ └── ...
│
├── vocalization/
│ ├── vocal001.wav
│ └── ...
│
├── background/
│ ├── fan001.wav
│ ├── tv001.wav
│ └── ...
│
└── other_noise/
├── door001.wav
└── ...
Recommended process :
Collect audio
↓
Label audio
↓
Remove unusable recordings
↓
Split train/validation/test
↓
Extract features
↓
Train model
↓
Evaluate
↓
Quantize/compress
↓
Deploy to ESP32
Do not collect or upload recordings of real babies without appropriate consent, privacy safeguards, and compliance with applicable requirements.
33. AI Model Evaluation
Do not report only accuracy.
Measure:
-
Precision
-
Recall
-
F1 score
-
False-positive rate
-
False-negative rate
-
Confusion matrix
-
Inference time
-
RAM consumption
-
Flash usage
Example confusion matrix:
Predicted
Cry Noise
Actual Cry TP FN
Actual Noise FP TN
For a baby-monitoring prototype, false alarms and missed events should both be explicitly evaluated rather than hiding them behind a single accuracy number.
34 . Safety Architecture
The cradle should have hardware safety mechanisms independent of the AI agent.
AI Agent
│
▼
Suggested action
│
▼
Safety controller
/ \
ALLOW BLOCK
│ │
▼ ▼
Limited action Stop
Recommended safeguards:
-
Physical motor limit
-
Maximum rocking duration
-
Emergency stop
-
Independent power cutoff
-
Watchdog timer
-
Motor driver protection
-
No unrestricted AI control
-
Local manual override
-
Sensor-failure detection
-
Wi-Fi failure fallback
Most importantly, the prototype should not rely on automated rocking or software to guarantee infant safety.
35. Power Architecture
AC/DC Adapter
│
┌─────┴─────┐
│ │
▼ ▼
ESP32 Motor
│
┌────────┼─────────┐
▼ ▼ ▼
DHT22 MPU6050 Microphone
Keep motor power and logic power appropriately isolated/decoupled, and ensure the selected supply, wiring, driver and enclosure are electrically and mechanically appropriate.
36. Project Software Stack
Firmware
└── Arduino IDE / PlatformIO
└── ESP32 C/C++
AI
└── TensorFlow Lite Micro /
Edge ML framework
Automation
└── n8n
Cloud
├── Google Sheets
└── ThingSpeak
Notification
└── Telegram
Dashboard
├── HTML
├── CSS
└── JavaScript
Optional backend
└── Node.js / Python / n8n API
37. Recommended Development Sequence
Phase 1 – ESP32
First make the ESP32 read:
Temperature
Humidity
Motion
Phase 2 – Wi-Fi
Confirm:
ESP32 → Wi-Fi → Internet
Phase 3 – n8n
Create:
ESP32 → n8n Webhook
Phase 4 – Google Sheets
Add:
n8n → Google Sheets
Phase 5 – ThingSpeak
Add:
n8n → ThingSpeak
Phase 6 – Telegram
Add:
n8n → Telegram
Phase 7 – Audio
Add:
Microphone → ESP32
Phase 8 – AI
Deploy:
Audio → ML model → Cry probability
Phase 9 – AI Agent
Add:
Sensor data → AI agent → structured event summary
Phase 10 – Dashboard
Add:
Cloud data → Web dashboard
Phase 11 – Testing
Test every subsystem independently before integrating the entire cradle.
38. Final Integrated Workflow
SMART CRADLE
│
┌────────────────────┼────────────────────┐
│ │ │
▼ ▼ ▼
Microphone DHT22 MPU6050
│ │ │
└────────────────────┼────────────────────┘
▼
ESP32
│
Edge processing
│
Cry probability
│
▼
Wi-Fi
│
▼
n8n Webhook
│
┌────────────────┼────────────────┐
│ │ │
▼ ▼ ▼
Google Sheets ThingSpeak AI Agent
│ │ │
│ │ ▼
│ │ Event interpretation
│ │ │
└────────────────┼────────────────┘
│
▼
Alert decision
│
▼
Telegram
/ \
Text Voice
│ │
└────┬────┘
▼
Parent
│
▼
Web Dashboard
39. Expected Final Demonstration
A project demonstration can follow this sequence:
-
Power on the smart cradle.
-
ESP32 connects to Wi-Fi.
-
Dashboard shows ONLINE.
-
Temperature and humidity appear.
-
Motion sensor reports cradle activity.
-
Audio is processed.
-
Cry classifier produces a probability.
-
A sustained crying event is detected.
-
ESP32 sends JSON to n8n.
-
n8n stores the event in Google Sheets.
-
n8n updates ThingSpeak.
-
AI agent interprets the event.
-
n8n sends Telegram text notification.
-
n8n generates a voice notification.
-
Parent receives the alert.
-
Dashboard changes to the current state.
-
Event remains available in the history.
40. Final Project Title
A suitable formal project title is:
“AI Smart Cradle with Baby Cry Detection, Sleep-State Analysis and Agentic IoT Automation Using ESP32, n8n, Telegram, Google Sheets and ThingSpeak.”
Alternative shorter title:
“AI-Powered Smart Cradle with ESP32 and Agentic IoT Automation.”
41. Key Innovation
The main innovation is not simply detecting sound. It is the integration of:
EDGE AI
+
IoT SENSORS
+
ESP32
+
n8n AUTOMATION
+
AI AGENT
+
CLOUD DATA
+
TELEGRAM VOICE ALERTS
+
WEB DASHBOARD
This creates an end-to-end agentic IoT monitoring platform in which the ESP32 performs local sensing, n8n coordinates events and services, the AI agent interprets events, and the caregiver receives actionable notifications.
The system should always be presented as an assistive prototype. Cry detection and sleep-state estimation are probabilistic and should not be represented as medical diagnosis or as a substitute for direct caregiver supervision.
Project Summary
AI Smart Cradle with Baby Cry Detection & Sleep Analysis is an IoT-based smart cradle using ESP32 + AI/ML + n8n + Telegram + Google Sheets + ThingSpeak + Web Dashboard.
Core workflow
Baby
↓
Microphone + Temperature + Humidity + Motion Sensors
↓
ESP32
↓
AI Cry Detection + Sleep/Activity Estimation
↓
Wi-Fi
↓
n8n Automation
├── Google Sheets → Data/Event Logging
├── ThingSpeak → Cloud Graphs
├── AI Agent → Event Analysis
└── Telegram → Text + Voice Alerts
↓
Parent
Main functions
-
🎤 Detect possible baby crying using an audio ML model.
-
🌡️ Monitor temperature and humidity.
-
📡 Monitor cradle/baby movement.
-
😴 Estimate states such as quiet sleep, active, or possible crying.
-
🤖 Use an AI agent to summarize events.
-
⚙️ Use n8n to automate the complete workflow.
-
📱 Send Telegram notifications.
-
🔊 Send Telegram voice alerts through a TTS service.
-
📊 Store historical data in Google Sheets.
-
📈 Visualize IoT data with ThingSpeak.
-
🌐 Provide a live web dashboard.
Hardware
-
ESP32 DevKit
-
I2S microphone
-
DHT22/SHT31
-
MPU6050
-
Vibration sensor
-
OLED display
-
LED/buzzer
-
Optional motor + motor driver
Software
ESP32
↓
Arduino IDE / PlatformIO
↓
Edge AI / TensorFlow Lite Micro
↓
n8n
├── Google Sheets
├── ThingSpeak
├── AI Agent
└── Telegram
├── Text
└── Voice
↓
Web Dashboard
Important design principle
The AI agent should not directly control the cradle motor without independent safety controls. Hardware limits, watchdogs, emergency stop/manual override, and bounded operating times should be implemented separately.
The complete project can therefore be presented as an end-to-end Agentic IoT system:
Sense → Analyze → Automate → Store → Notify → Visualize.
🧠 AI Smart Cr adle — Mind Map
┌──────────────────────────────┐
│ AI SMART CRADLE │
│ Baby Cry Detection + Sleep │
│ Analysis + Agentic IoT │
└──────────────┬───────────────┘
│
┌─────────────────────────────────────┼─────────────────────────────────────┐
│ │ │
▼ ▼ ▼
┌───────────────┐ ┌───────────────┐ ┌───────────────┐
│ HARDWARE │ │ AI/ML │ │ IoT │
└───────┬───────┘ └───────┬───────┘ └───────┬───────┘
│ │ │
├── ESP32 ├── Audio preprocessing ├── Wi-Fi
├── I2S Microphone ├── MFCC / Mel features ├── HTTP/JSON
├── DHT22 / SHT31 ├── Cry classifier ├── n8n
├── MPU6050 ├── Cry probability ├── Google Sheets
├── Vibration sensor ├── Sleep estimation ├── ThingSpeak
├── OLED ├── Activity detection └── Cloud API
├── LED
├── Buzzer
└── Optional Motor
│
▼
┌─────────────────────┐
│ EVENT ANALYSIS │
└──────────┬──────────┘
│
┌──────────────────┼──────────────────┐
│ │ │
▼ ▼ ▼
Normal Sleeping Crying
│ │ │
└──────────────────┼──────────────────┘
│
▼
┌─────────────────────┐
│ n8n AUTOMATION │
└──────────┬──────────┘
│
┌──────────────────────────┼─────────────────────────┐
│ │ │
▼ ▼ ▼
┌───────────────┐ ┌───────────────┐ ┌───────────────┐
│ Google Sheets │ │ ThingSpeak │ │ AI AGENT │
└───────────────┘ └───────────────┘ └───────┬───────┘
│ │ │
▼ ▼ ▼
Event Log IoT Graphs Event Summary
│
▼
┌────────────────────┐
│ ALERT DECISION │
└─────────┬──────────┘
│
┌──────────────┴─────────────┐
│ │
▼ ▼
Telegram Text Telegram Voice
│ │
└──────────────┬─────────────┘
▼
👨👩👧 Parent
│
▼
┌──────────────────┐
│ WEB DASHBOARD │
└────────┬─────────┘
│
┌──────────────────────────────┼───────────────────────┐
│ │ │
▼ ▼ ▼
Temperature Humidity Cry Probability
│ │ │
└──────────────────────────────┼───────────────────────┘
│
▼
Baby State
🔄 Core Concept
SENSE
↓
ESP32 Sensors
↓
PROCESS
↓
Edge AI / Cry Detection
↓
ANALYZE
↓
n8n + AI Agent
↓
AUTOMATE
↓
Google Sheets + ThingSpeak
↓
NOTIFY
↓
Telegram Text + Voice
↓
VISUALIZE
↓
Web Dashboard
🎯 Project Goal
“Build an intelligent, connected cradle that detects possible crying and activity locally, analyzes events through an automated AI/IoT pipeline, stores historical data, and provides timely caregiver notifications.”
No comments:
Post a Comment