Robust Analysis of Motor Imagery From Brain Signals for a BCI-Controlled Virtual Reality System to Aid Paralysis Patients

2. Previous Research Study

Utilization of brain signals is the utmost significant issue for controlling BCI technology. There are many kinds of recorded brain activity, i.e. MEG (Kheirkhah et al., 2020; Tait et al. 2021; Chholak et al., 2021; Caffarra et al. 2021; Hamilton et al., 2021), EEG (Wang et al., 2014; Alzhrani et el., 2021; Want et al., 2021; Gao et al. 2021; Song et al. 2021), ECoG (Liu et el., 2020; Semenova et al., 2021, Ajrawi et al., 2021; Uehara et al., 2021; Krishnan and Bai, 2021); Frolov et al., 2017). MEG data represents the magnetic of the brain and provides less noisy data (Kheirkhah et al., 2020) shown in Fig. 1. However, MEG data is expensive and risky to record in real time. In this context, ECoG data are the most unsafe to acquire in case if the intention to use this type of data in invasive technology (Liu et al., 2020). Although ECoG data is not flexible to collect, ECoG data provides low amount of noise signals than EEG data. Both, EEG and ECoG data were used previously to record motor imagery activities of human brain, MEG data was normally used for studying human facial expression. EEG data are the safest and easiest to acquire from any other brain signals due to acquire these data by conductive gel and electrode cap during data acquisition (Wang et al., 2014). In this context, it is mentionable that EEG data has been using by existing research instead of ECoG data due to less risk of noninvasive data acquisition characteristics. However, EEG datasets are very much noisy which requires cleaning before classifying during features extraction.