Brain-computer interface (BCI) has attracted attention as a new type of interface for both healthy and health-challenged individuals. Many BCI studies using steady-state visual evoked potential (SSVEP) have been reported. However, the conventional SSVEP-based BCI was found difficult to use as a versatile interface in real-life settings because the system included several types of apparatuses and computers that are not easily portable. Therefore, we investigated a new SSVEP-based BCI, our system included smart glasses for visual stimuli and employed a mobile neuro-headset for measuring electro-encephalogram (EEG) signals. We conducted experiments using a two-class SSVEP with our system. An SSVEP dataset from five healthy subjects was used in averaging classification performance. The classification performance depended on the viewing area of pattern reversal on the smart glasses was estimated. The accuracy rate was high when a large pattern was used. Next, the classification performance in three different situations -- a resting state, a walking state, and a conversation state -- was measured. Classifiers using features up to the second harmonic SSVEP signals gave a mean accuracy rate of 0.70 at the resting state.