Speech imagery based vowel classification from eeg signals using machine learning techniques

Abstract

Speech is a complex brain activity that includes a series of actions newlineperformed by various brain regions. Neurological speech disorder is a newlinesubcategory of speech disorder that causes loss of speech due to the partial or newlineincomplete neural connections between the regions of the brain involved in newlinespeech processes such as comprehension and production. The comprehended newlinespeech can be replayed inside the brain as covert speech or mind voice or newlineeven called speech imagery. Speech Imagery (SI) is imagining speaking a newlinesegment of speech. Decoding the SI means unwrapping the brain changes newlineduring the speech imagery process and using these brain signatures and aids newlinein building speech-based neural prosthetics using artificial intelligence. The newlineimagined speech can be decoded using EEG signals recorded while imagining newlinespeaking. A lot of research has been carried out in decoding the imagined newlinespeech. Decoding vowels from Electroencephalographic signals can be of newlinesubstantial relevance in identifying the correct words in many Brain Computer newlineInterface (BCI) applications and is also a dependable support system for people newlinewith neurological speech disorders. Though identifying vowels has been studied newlinefor a long time, the most difficult part is deciding on the right signal-processing newlineroutines and classifiers. This work aims to use several types of machine learning newlinetechniques to decode the imagined speech and incorporate technologies to newlineenhance performance and compare performance to create a better newlineneuroprosthetics device that will allow people with speech disorders due to newlineneurological impairments to communicate naturally. newline