An Intelligent Companion Robot Based on Visual and Speech Recognition

With the rise of super-aged societies, elderly care is a critical issue. To reduce injury risks and enhance user engagement, we have developed an intelligent companion robot integrating visual and voice recognition. The key functions include object retrieval, autonomous following, and interactive frisbee play. Modes can be switched via voice command or a mobile app. The robot's head and tail collaboration to perform the picking up function for the elderly, thereby reducing the risk of injury. While several companion robots exist, they typically lack the crucial function of assisting with object retrieval. The system utilizes YOLOv8 for object recognition, LiDAR for obstacle avoidance, and Google Gemini for AI voice interaction. Experimental results show an indoor object recognition success rate of approximately 90 %, while different model formats can enhance inference speed by 50 %. These results verify the effectiveness of the system functionality for elderly care.