Abstract:Objective. To evaluate the potential of intracortical electrode array signals for brain-computer interfaces (BCIs) to restore lost speech, we measured the performance of decoders trained to discriminate a comprehensive basis set of 39 English phonemes and to synthesize speech sounds via a neural pattern matching method. We decoded neural correlates of spoken-out-loud words in the 'hand knob' area of precentral gyrus, a step toward the eventual goal of decoding attempted speech from ventral speech areas in patients who are unable to speak. Approach. Neural and audio data were recorded while two BrainGate2 pilot clinical trial participants, each with two chronically-implanted 96-electrode arrays, spoke 420 different words that broadly sampled English phonemes. Phoneme onsets were identified from audio recordings, and their identities were then classified from neural features consisting of each electrode's binned action potential counts or high-frequency local field potential power. Speech synthesis was performed using the 'Brain-to-Speech' pattern matching method. We also examined two potential confounds specific to decoding overt speech: acoustic contamination of neural signals and systematic differences in labeling different phonemes' onset times. Main results. A linear decoder achieved up to 29.3% classification accuracy (chance = 6%) across 39 phonemes, while an RNN classifier achieved 33.9% accuracy. Parameter sweeps indicated that performance did not saturate when adding more electrodes or more training data, and that accuracy improved when utilizing time-varying structure in the data. Microphonic contamination and phoneme onset differences modestly increased decoding accuracy, but could be mitigated by acoustic artifact subtraction and using a neural speech onset marker, respectively. Speech synthesis achieved r = 0.523 correlation between true and reconstructed audio. Significance. The ability to decode speech using intracortical electrode array signals from a nontraditional speech area suggests that placing electrode arrays in ventral speech areas is a promising direction for speech BCIs.

Chisco: an EEG-based BCI Dataset for Decoding of Imagined Speech

ChineseEEG: A Chinese Linguistic Corpora EEG Dataset for Semantic Alignment and Neural Decoding

Simulation Experiment of BCI Based on Imagined Speech EEG Decoding

Electroencephalogram (EEG) Based Imagined Speech Decoding and Recognition

Speech decoding from stereo-electroencephalography (sEEG) signals using advanced deep learning methods

Imagined speech can be decoded from low- and cross-frequency intracranial EEG features

A State-of-the-Art Review of EEG-Based Imagined Speech Decoding

Decoding Covert Speech From EEG-A Comprehensive Review

Continuous and discrete decoding of overt speech with scalp electroencephalography (EEG)

Continuous and discrete decoding of overt speech with electroencephalography

Decoding High-level Imagined Speech using Attention-based Deep Neural Networks

Translating Mental Imaginations into Characters with Codebooks and Dynamics-Enhanced Decoding

Resting state EEG assisted imagined vowel phonemes recognition by native and non-native speakers using brain connectivity measures

The Potential for a Speech Brain–Computer Interface Using Chronic Electrocorticography

“Thinking out loud”: an open-access EEG-based BCI dataset for inner speech recognition

Inferring imagined speech using EEG signals: a new approach using Riemannian manifold features

Thinking out loud, an open-access EEG-based BCI dataset for inner speech recognition

Decoding Imagined Speech and Computer Control using Brain Waves

Decoding spoken English from intracortical electrode arrays in dorsal precentral gyrus

Silent EEG Classification Using Cross-Fusion Adaptive Graph Convolution Network for Multilingual Neurolinguistic Signal Decoding.