I. Introduction
The emergence of Internet of Things (IoT) has been extensively making a huge impact on human lives. Today, many IoT applications have been built for smart cities, smart homes, smart health care and vehicular networks [1]. Apart from these inexhaustive benefits, the field of cognitive neuroscience have also been inspired by the numerous advantages of connecting the brain to the cloud via the internet [2]. On several occasions, the Electroencephalography (EEG) based Brain-Computer Interface (BCI) has been widely explored due to its non-invasiveness and portability when integrating with the internet [3]. For example, the translation of human intentions into visible actions over the internet [4], and the IoT enabled thought-controlled wheelchair [5], are the recent trends in EEG based BCI research which is suggested to be the major technology that will support brain-to-thing interaction [6]. EEG based BCI offers variety of applications in the medical field. One of these is the Rehabilitation and restoration of sensory functions in stroke patients [7]. In which the impaired motor functions are restored through neuroplasticity. Another is the use of neuroprosthesis or prosthetic limb to regain lost mobility and functions in amputees [8]. Furthermore, the application of motor imagery (MI) signals from EEG-based BCI are equally significant in neurofeedback systems for motor therapy in post stroke patient [9]. However, the EEG-based BCI are faced with a number of challenges, which are primarily susceptible to a very low signal-to-noise ratios, easily influenced by environmental factors, and inherently lack sufficient spatial resolution [10]. In addition, preprocessing of EEG signals, the feature extraction and selection techniques are time-consuming and strongly rely on human expertise in the field [11]. To address these issues, we proposed an extended particle swarm optimization (PSO) based on neural network (NN) which selects appropriates features to be trained for accurate predictions. Consequently, the application of this technique will efficiently improve interaction between the brain and IoT device.