Starting with Machine Learning on devices like MCU - Part 1

Mostly Machine Learning is used on MCU when we think of Edge Computing. Instead of sending data to the cloud, analytics can be performed at the device level itself.

To start with machine learning on MCU, you need peripheral devices that:

• Gather data (sensors)
• Process or store data (memory, accelerators)
• Communicate results (outputs or interfaces)

The below sensors can be thought of if you are planning to start Machine Learning at MCU level.

| **ML Application**          | **Useful Sensors**                     | **Examples**                                  |

|-----------------------------|----------------------------------------|--------------------------------------------------|

| Gesture / motion detection  | Accelerometer, Gyroscope               | MPU6050, LSM6DS3                 |

| Sound / voice recognition   | Microphone                             | INMP441, SPH0645                             |

| Image / object detection    | Camera module                          | OV7670, Himax HM01B0                   |

| Environmental monitoring    | Temperature, Humidity, Gas, Light  | DHT11, BME280, BH1750       |

| Predictive maintenance      | Vibration, Current, Magnetic           | ADXL345, ACS712, MLX90393  |

Users can start with a MCU like,

• Arduino Nano 33 BLE Sense
• Arduino Uno R4 WiFi board
• ESP32
• STM32

Some of the tiny models such as the below can be run:

• Gesture recognition
• Keyword spotting ("Yes/No")
• Environmental anomaly detection