Please or Register to create posts and topics.

Request for EEG Datasets Related to Imagined Speech Research

Dear Team,

Thank you in advance for your assistance.

I am currently searching for EEG datasets that include signals from subjects with aphasia or other conditions, specifically oriented towards research on imagined speech. My goal is to process these EEG signals to reproduce the words or phrases of a subject who cannot emit sounds but can think about what they want to say.

I understand that aphasia and conditions like ALS may involve different processing of speech, and I am investigating these differences. My initial step is to obtain relevant EEG datasets to develop deep learning models for this purpose. I have previously worked with models to process EEG signals and identify distinctive patterns between younger and older subjects, examining cognitive dysfunction related to aging and the associated structural and functional imbalances in brain activity.

If you have any recommendations or can provide access to datasets that align with this research focus, I would greatly appreciate your help.

Best regards,

Thank you so much for your interest, it is really a very specific area, we are working to create a platform with an EEG dataset, but in the first step, it will be for general tasks such as signal processing and feature extraction.

It's a very interesting field that can help a lot of people. On my GitHub, I have a repository (MyRC) where I conduct a study on EEG signal processing using RC-ESN. I provide an API or framework for both the preprocessing of these signals and their subsequent processing with the RC model. Specifically, it focuses on studying the aging process and identifying distinctive patterns between young, adult, and older subjects. It corroborates all the hypotheses and shows structural and cognitive functional deterioration with aging. Additionally, it reveals that each subject's brain activity is unique, with greater heterogeneity among younger individuals due to higher plasticity and diverse studies and experiences during growth.

The website is: https://github.com/jogugil/MyRC

It is in Spanish, but if you're interested, I could look into translating parts into English.

I am currently examining which parameters are most important, using optimization methods to obtain hyperparameter values, which is quite challenging, and determining which ML models better utilize the information returned by RC-ESN.

The advantage of RC-ESN is that it eliminates the need for real-time feature extraction in time, frequency, etc. However, the framework does include a library and function where you can provide an EEG signal dataset, and it will return a three-dimensional matrix for each subject with features per EEG signal.

 

Thank you so much, looks really great!  Will be in touch!