Table 3.

Imagined speech classification methods summary (continuation).

References	Task	Methods	Brain area and waves	Performance
Arjestan et al. (2016)	Binary classification vowels, syllables, and resting state	- CSP	- All areas	Vowels: 76.6% best accuracy
		- SVM	− 8–45 Hz bandpass	Syllables: 76.4% best accuracy
Sereshkeh et al. (2017)	Binary classification of “yes,” “no,” and rest state	- Multilayer perceptron	- All areas	70% Average accuracy
		- DWT	− 0–50 Hz bandpass
		- RMS / SD
Nguyen et al. (2017)	Classification of vowels, short, and long words	- Riemannian Manifold	- Broca's area, the motor cortex and Wernicke's area	- Vowels: 49% Average accuracy
		- RVM	− 8–70 Hz bandpass	- Short words: 50.1% Average accuracy
		- CSP / WT		- Long words: 66.2% Average accuracy
				- S-L: 80.1% Average accuracy
Paul et al. (2018)	Classification of three Hindi vowels	- SVM	- Broca's and Wernicke's area	63% Average accuracy
		- AR coefficients	− 0.1–36 Hz bandpass
		- Hjorth parameters
		- Sample entropy
Cooney et al. (2018)	Classification of seven phonemic prompts and four words	- SVM	- All areas	- Phonemes: 20% Average accuracy
		- Decision Tree	− 1–50 Hz bandpass	- Pair of words: 44% Average accuracy
		- MFCC
		- ICA