Slimming Edge AI: A Case Study of Gesture Recognition for Air Conditioner Control

The increasing popularity of machine learning (ML)/artificial intelligence (AI) applications in smart homes and medical care has sparked interest in edge/local AI. This study focuses on gesture-based air conditioner control, comparing two popular microcontroller-based AI architectures: TinyML and AI for embedded systems framework (AIfES). Our experimental results show that AIfES significantly outperforms TinyML (using TensorFlow Lite for microcontroller, TFLM) in deployment resources. This suggests that the use of lightweight frameworks like AIfES on low-cost microcontrollers for simple applications can greatly enhance cost-effectiveness, paving the way for future research and development in this field.