Seamless human-device interaction in the internet of things

Abstract

The Internet of Things will bring a scenario in which interaction between humans and devices will be critical to allow people to use, monitor or configure Internet of Things devices. Interactions in such applications are based on traditional graphical interfaces. Devices that accept interaction based on Natural Language, e.g., through voice commands, can understand basic human orders or answering questions whenever user expressions fit into a known language pattern. Some devices can understand natural language voice commands but require sophisticated voice assistants located in the cloud, which raises significant privacy concerns. Others devices which handle voice-processing locally can perform a very limited local recognition system, requiring users to be familiar with words the system can process. The purpose of this work is to diminish the complexity of Natural Language processing in the context of IoT. The solution posited in this article allows Internet of Things devices to offload Natural Language processing to a system that improves the use of Natural Language and alleviates the need to learn or remember specific words or terms intended for triggering device actions. We have evaluated the feasibility of the design with a proof-of-concept implemented in a home environment and it was tested by real users.