A LoRa-Based Emotion Estimation Scheme for Smart Home Automated Actions Using ELMs

Emotion detection is crucial in many IoT deployments from an operational perspective with examples ranging from digital health to smart cities. This is particularly true in smart homes where the interaction between the local IoT ecosystem and the inhabitants are continuous, pervasive, and nuanced. More specifically, emotion estimation from human speech attributes is an integral architectural component of such ecosystems. In this work, we survey the emerging LPWAN technologies and after selecting the most optimal for our use-case, we propose an emotion estimation scheme based on LoRa wireless technology for automated actions in smart home environments. In particular, a voice recognition module coupled with a transmitter are installed in a car. Then, depending on the estimation outcome, the smart home may undertake one or more preemptive actions according to its configuration as the car passenger approaches. The prototype LoRa system has been tested with the widely-used TESS dataset with encouraging results as expressed in the emotion confusion matrix obtained from a extreme learning machine with various kernels. The preceding paves the way for adaptive smart homes tailored to the needs of their inhabitants.