A new study published in the Journal of Neural Engineering this month demonstrates how a brain-computer interface (BCI) utilizes artificial intelligence (AI) deep learning to convert brain activity into speech with an accuracy rate of up to 100%.
"This study shows that high accuracy and robust decoding can be achieved even with a relatively small dataset (10 repetitions of 12 words) when using speech reconstruction for classification," wrote Julia Berezutskaya, the lead author of the study and a postdoctoral researcher at the Donders Institute for Brain, Cognition, and Behavior at Radboud University. "These results highlight the potential of this approach for further use in BCI."
Brain-computer interfaces, also known as brain-machine interfaces (BMI), offer hope for individuals who have lost the ability to speak or move. They operate and control robotic limbs, computer software applications (such as email), and other external devices by decoding the intentions of patients from their brain activity.
"So far, there has been no comprehensive study on optimizing deep learning models for speech reconstruction. Additionally, there is a lack of consensus on the selection of brain and audio speech features used in such models," the researchers wrote.
The team utilized high-density electrocorticography (HD ECoG) grids implanted in the somatosensory motor cortex region, which generates brain activity during the speech production process, for speech reconstruction. They validated and improved the neural decoding methods used in this study.
"Understanding which decoding strategies can provide the best and most directly applicable results is crucial for advancing this field," the scientists wrote.
Speech reconstruction uses brain activity data as input to generate a graphical representation called a spectrogram. The participants in the study were implanted with the NeuroPort neural recording system from Blackrock Microsystems, which collected brain activity data from the somatosensory motor region when they loudly spoke 12 words ten times each.
The researchers evaluated three different deep learning speech reconstruction models: a sequence-to-sequence (S2S) recurrent neural network (RNN), a multilayer perceptron (MLP), and a DenseNet (DN) convolutional neural network (CNN).
The artificial intelligence machine learning classifiers achieved decoding accuracies ranging from 92% to 100% for individual words in reconstructed speech across all models. Furthermore, they found that highly complex artificial intelligence deep neural network models were required for more accurate artificial intelligence speech reconstruction.
The multilayer perceptron (MLP) had a relatively simple computational architecture consisting of basic linear operations and nonlinear activation functions, and its performance surpassed that of artificial intelligence models with more complex computations. The recurrent sequence-to-sequence with attention mechanism and state memory, as well as the convolutional DenseNet with skip connections and local convolutions, both utilized more complex computations than the MLP AI model.
The research findings indicate that combining artificial intelligence and brain-machine interfaces to directly reconstruct speech from brain activity in the somatosensory motor region can achieve highly accurate word decoding.
The research team concluded, "These results have the potential to advance state-of-the-art techniques in speech decoding and reconstruction for future communication in BCI."
Alphabet Q1 paid subscribers reached 350M, fueled by YouTube & Google One. YouTube ad revenue slightly missed at $9.88B amid Premium growth. Total Q1 revenue hit $109.9B, beating forecasts.
Anthropic releases Claude connectors for Adobe, Blender, and Ableton, enabling AI workflow automation and app control. It also sponsors the Blender Development Foundation.
Apple introduces an App Store model splitting 12-month plans into monthly installments. On iOS 26.4+, it lowers upfront costs without reducing total price, adding billing tracking.
Google launches Aadhaar digital IDs in Wallet for India, enabling secure QR/NFC verification. A phased partner rollout targets key sectors, while passport-based credentials expand globally.
YouTube tests "Ask YouTube," an AI tool blending text and video for complex queries. Initially for US Premium users, it supports interactive follow-ups, expanding Google's AI search.
Google grants US DoD classified AI access, joining OpenAI and xAI in defense deals amid Anthropic's dispute. Contractual safeguards against surveillance and lethal weapons remain unenforceable.
Mistral AI launches Workflows, a Temporal-powered orchestration platform for secure, Python-driven enterprise AI automation featuring human-in-the-loop approvals.