Smart home technology has received growing interest in recent years with a focus on automation and assistance, for example, Alexa, Google Home, Apple HomePod, and many cheap IoT devices. Better supporting elderly people to continue live in their home using smart home technology is a key target application. However, most of the existing smart home solutions for the elderly are not designed with people's emotional goals in mind, leading to lack of adoption, lack of engagement, and failure of the technology. In this paper, we introduce an emotion-oriented requirements engineering approach to help identifying, modeling and evaluating emotional goals. We also explain how we used this technique to help us develop SofiHub - a new smart home platform for elderly people. SofiHub comprises a range of devices and software for sensing, interaction, passive monitoring, and emergency assistance. We have conducted multiple trials including initial field trials for elderly people in real houses. We have used our emotion-oriented requirements techniques to evaluate the participants’ emotional reactions before, during, and after trials to understand the impact of such technology on elderly people's emotions to the SofiHub solution. Our analysis shows that SofiHub successfully alleviates their loneliness, makes them feel safer and cared about. We also found that the trial participants developed a strong relation with the system and hence, felt frustrated when SofiHub did not respond in ways expected or desired. We reflect on the lessons learned from the trials related to our emotion-oriented design and evaluation experimental approach, including refining our set of evaluation tools.