Brain computer interfaces (BCI) are the fledgling field to rehabilitate the immobilized people. This BCI technology can succour paralyzed patients to operate wheelchairs independently for locomotion, also to lift and carry the objects based on brain-neuronal activity with robotic control. EEG (Electroencephalography) is a device used to provide information immeasurably identifying about brain conditions and disabilities with an effective stimulus using graphite or noble metals. It indicates extremely herculean and targeted EEG applications to guide devices utilizing brain activity. This study gives the survey of (ML) machine learning and (DL) deep learning associated with MI (Motor Imagery), MeI (Mental Imagery) and ME (Motor Execution) gesture classifications applicable for BCI. There are two public domain datasets (PhysioNet, BCIC- Motor Execution) and self-collected datasets were utilized for computerized process since inception. DBN (Deep belief networks), PCA (Principal component analysis) and few other transforms were applied to the recorded signal to extract the features. To categorize the features derived from the collected signal, diverse machine learning and deep learning classification algorithms are available. This article surveys on literature for multiplex weighted visibility graph (MWVG), G-CRAM (Graph based convolutional Re current attention model), FFNN (Feedforward neural network classifier), HF-CNN (Hierarchical flow conventional neural network), (Spectro temporal Decomposition – Squeeze and Excitation – Convolutional Neural Network) SSD-SE-CNN, LASSO (Least absolute shrinkage and selection operator), DML (Deep metric learning), (ASTGCN) adaptive spatiotemporal graph convolutional network, VaS-LDA (Vertical arrangements of sub-bands - Linear Discriminant Analysis), ZSL (Zero shot learning), RLS-CSP (Recursive-least squares updates of the Common Spatial Pattern filter coefficient), SW-LCR (Sliding window- Longest consecutive repetition), CNN (Convolutional Neural Networks), SVM-RBF (Radial Basis Function-based support vector machine (SVM) classifier), DJDAN (dynamic joint domain adaptation network), and S-EEGNet (Separable EEGNet) with HHT (Hilbert-Huang Transform). Eventually, this work also consolidates wide range of research progress in classification and analysis on the datasets, sampling rate, number of subjects and overall performances are discussed in specific to motor imaginary tasks.