Machine-Learning Methods for Speech and Handwriting Detection Using Neural Signals: A Review.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Brain-Computer Interfaces (BCIs) offer transformative potential for individuals with severe motor and communication disabilities.
• Advancements in decoding neural signals for speech and handwriting can create accessible communication platforms.
• BCIs are expanding beyond medical applications into cognitive training, gaming, and AR/VR.

Purpose of the Study:

• To systematically review and analyze existing research on neural signal-based handwriting and speech recognition.
• To provide a comprehensive overview of methodologies, datasets, and preprocessing techniques for new researchers in the field.
• To consolidate knowledge on both invasive and non-invasive BCI approaches for text generation from neural data.

Main Methods:

• Categorization of current research into invasive and non-invasive BCI studies.
• Examination of papers focusing on converting speech- and handwriting-activity-based neural signals into text.
• Discussion of data extraction methods from the brain and analysis of published datasets (2014-2022).

Main Results:

• Identified two primary categories of neural signal-based recognition: invasive and non-invasive.
• Detailed the methodologies for extracting and processing neural data for speech and handwriting decoding.
• Summarized key datasets, preprocessing steps, and machine learning techniques employed in recent literature.

Conclusions:

• Neural signal-based recognition holds significant promise for enhancing communication for individuals with disabilities.
• A comprehensive understanding of current methodologies is crucial for advancing BCI technology.
• This review serves as a foundational resource for researchers exploring machine learning applications in neural signal processing for communication.