TinyML-based anomaly detection

Tiny machine learning (TinyML) is a burgeoning field at the intersection of embedded systems, machine learning, and performance engineering. It empowers ultra-low power devices like MCU to infer original data at the endpoint with an ML algorithm. On-device inferring relieves the omnipresent dependency on remote servers. It dispels the barriers of traditional resource-intensive ML and benefits low-power operation, economics, low latency, and privacy protection. The efficient TinyML technology brings more intelligence to the battery-powered and always-on endpoint devices, which will definitely unleash a plethora of applications in the future.

The simple demos above are based on tinyml for vibration analysis. The vibration signals are colected and inferred at the endpoint. The first one is based on the commonly utilized IMU (Inertial Measurement Unit), and the second one is based on a self-powered sensor, which can cut the energy consumption while sensing. These projects improved my development experience for edge intelligence.