SmartPose: An Intelligent Household Appliances Controller Based on Visual Recognition of Human Postures

Smart home can greatly increase the convenience of life, and has become a research hotspot nowadays. The vision-based smart home controlling system is widely used in rooms, such as kitchens, where the needs to pressing switches or touching screens can cause many inconveniences. This paper proposes an intelligent household appliances controller based on visual recognition for human posture, namely SmartPose. This system employs the Back Propagation Neural Network to recognize the key points of the human joints' key points extracted by a human joints recognition library, namely OpenPose, and then generates posture results. After that, a Raspberry Pi receives the results and controls the home appliances through a relay. The experimental results indicate that this system's accuracy for recognizing five postures is 99.8%, which can be applied to the control of household appliances.