Automatic Speech Recognition from Neural Signals: A Focused Review

Speech synthesis from three-axis accelerometer signals using conformer-based deep neural network

Silent speech interfaces (SSIs) have emerged as innovative non-acoustic communication methods, and our previous study demonstrated the significant potential of three-axis accelerometer-based SSIs to identify silently spoken words with high classification accuracy. The developed accelerometer-based SSI with only four accelerometers and a small training dataset outperformed a conventional surface electromyography (sEMG)-based SSI. In this study, motivated by the promising initial results, we investigated the feasibility of synthesizing spoken speech from three-axis accelerometer signals. This exploration aimed to assess the potential of accelerometer-based SSIs for practical silent communication applications. Nineteen healthy individuals participated in our experiments. Five accelerometers were attached to the face to acquire speech-related facial movements while the participants read 270 Korean sentences aloud. For the speech synthesis, we used a convolution-augmented Transformer (Conformer)-based deep neural network model to convert the accelerometer signals into a Mel spectrogram, from which an audio waveform was synthesized using HiFi-GAN. To evaluate the quality of the generated Mel spectrograms, ten-fold cross-validation was performed, and the Mel cepstral distortion (MCD) was chosen as the evaluation metric. As a result, an average MCD of 5.03 ± 0.65 was achieved using four optimized accelerometers based on our previous study. Furthermore, the quality of generated Mel spectrograms was significantly enhanced by adding one more accelerometer attached under the chin, achieving an average MCD of 4.86 ± 0.65 (p < 0.001, Wilcoxon signed-rank test). Although an objective comparison is difficult, these results surpass those obtained using conventional SSIs based on sEMG, electromagnetic articulography, and electropalatography with the fewest sensors and a similar or smaller number of sentences to train the model. Our proposed approach will contribute to the widespread adoption of accelerometer-based SSIs, leveraging the advantages of accelerometers like low power consumption, invulnerability to physiological artifacts, and high portability.

Neural Decoding of Spontaneous Overt and Intended Speech

PURPOSE

The aim of this study was to decode intended and overt speech from neuromagnetic signals while the participants performed spontaneous overt speech tasks without cues or prompts (stimuli).

METHOD

Magnetoencephalography (MEG), a noninvasive neuroimaging technique, was used to collect neural signals from seven healthy adult English speakers performing spontaneous, overt speech tasks. The participants randomly spoke the words yes or no at a self-paced rate without cues. Two machine learning models, namely, linear discriminant analysis (LDA) and one-dimensional convolutional neural network (1D CNN), were employed to classify the two words from the recorded MEG signals.

RESULTS

LDA and 1D CNN achieved average decoding accuracies of 79.02% and 90.40%, respectively, in decoding overt speech, significantly surpassing the chance level (50%). The accuracy for decoding intended speech was 67.19% using 1D CNN.

CONCLUSIONS

This study showcases the possibility of decoding spontaneous overt and intended speech directly from neural signals in the absence of perceptual interference. We believe that these findings make a steady step toward the future spontaneous speech-based brain-computer interface.

Flexible Self-Powered Low-Decibel Voice Recognition Mask

In environments where silent communication is essential, such as libraries and conference rooms, the need for a discreet means of interaction is paramount. Here, we present a single-electrode, contact-separated triboelectric nanogenerator (CS-TENG) characterized by robust high-frequency sensing capabilities and long-term stability. Integrating this TENG onto the inner surface of a mask allows for the capture of conversational speech signals through airflow vibrations, generating a comprehensive dataset. Employing advanced signal processing techniques, including short-time Fourier transform (STFT), Mel-frequency cepstral coefficients (MFCC), and deep learning neural networks, facilitates the accurate identification of speaker content and verification of their identity. The accuracy rates for each category of vocabulary and identity recognition exceed 92% and 90%, respectively. This system represents a pivotal advancement in facilitating secure and efficient unobtrusive communication in quiet settings, with promising implications for smart home applications, virtual assistant technology, and potential deployment in security and confidentiality-sensitive contexts.

Online speech synthesis using a chronically implanted brain-computer interface in an individual with ALS

Brain-computer interfaces (BCIs) that reconstruct and synthesize speech using brain activity recorded with intracranial electrodes may pave the way toward novel communication interfaces for people who have lost their ability to speak, or who are at high risk of losing this ability, due to neurological disorders. Here, we report online synthesis of intelligible words using a chronically implanted brain-computer interface (BCI) in a man with impaired articulation due to ALS, participating in a clinical trial (ClinicalTrials.gov, NCT03567213) exploring different strategies for BCI communication. The 3-stage approach reported here relies on recurrent neural networks to identify, decode and synthesize speech from electrocorticographic (ECoG) signals acquired across motor, premotor and somatosensory cortices. We demonstrate a reliable BCI that synthesizes commands freely chosen and spoken by the participant from a vocabulary of 6 keywords previously used for decoding commands to control a communication board. Evaluation of the intelligibility of the synthesized speech indicates that 80% of the words can be correctly recognized by human listeners. Our results show that a speech-impaired individual with ALS can use a chronically implanted BCI to reliably produce synthesized words while preserving the participant's voice profile, and provide further evidence for the stability of ECoG for speech-based BCIs.

Direct speech reconstruction from sensorimotor brain activity with optimized deep learning models

Objective.Development of brain-computer interface (BCI) technology is key for enabling communication in individuals who have lost the faculty of speech due to severe motor paralysis. A BCI control strategy that is gaining attention employs speech decoding from neural data. Recent studies have shown that a combination of direct neural recordings and advanced computational models can provide promising results. Understanding which decoding strategies deliver best and directly applicable results is crucial for advancing the field.Approach.In this paper, we optimized and validated a decoding approach based on speech reconstruction directly from high-density electrocorticography recordings from sensorimotor cortex during a speech production task.Main results.We show that (1) dedicated machine learning optimization of reconstruction models is key for achieving the best reconstruction performance; (2) individual word decoding in reconstructed speech achieves 92%-100% accuracy (chance level is 8%); (3) direct reconstruction from sensorimotor brain activity produces intelligible speech.Significance.These results underline the need for model optimization in achieving best speech decoding results and highlight the potential that reconstruction-based speech decoding from sensorimotor cortex can offer for development of next-generation BCI technology for communication.

A Neural Speech Decoding Framework Leveraging Deep Learning and Speech Synthesis

Decoding human speech from neural signals is essential for brain-computer interface (BCI) technologies restoring speech function in populations with neurological deficits. However, it remains a highly challenging task, compounded by the scarce availability of neural signals with corresponding speech, data complexity, and high dimensionality, and the limited publicly available source code. Here, we present a novel deep learning-based neural speech decoding framework that includes an ECoG Decoder that translates electrocorticographic (ECoG) signals from the cortex into interpretable speech parameters and a novel differentiable Speech Synthesizer that maps speech parameters to spectrograms. We develop a companion audio-to-audio auto-encoder consisting of a Speech Encoder and the same Speech Synthesizer to generate reference speech parameters to facilitate the ECoG Decoder training. This framework generates natural-sounding speech and is highly reproducible across a cohort of 48 participants. Among three neural network architectures for the ECoG Decoder, the 3D ResNet model has the best decoding performance (PCC=0.804) in predicting the original speech spectrogram, closely followed by the SWIN model (PCC=0.796). Our experimental results show that our models can decode speech with high correlation even when limited to only causal operations, which is necessary for adoption by real-time neural prostheses. We successfully decode speech in participants with either left or right hemisphere coverage, which could lead to speech prostheses in patients with speech deficits resulting from left hemisphere damage. Further, we use an occlusion analysis to identify cortical regions contributing to speech decoding across our models. Finally, we provide open-source code for our two-stage training pipeline along with associated preprocessing and visualization tools to enable reproducible research and drive research across the speech science and prostheses communities.

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

Brain-Computer Interfaces (BCIs) have become increasingly popular in recent years due to their potential applications in diverse fields, ranging from the medical sector (people with motor and/or communication disabilities), cognitive training, gaming, and Augmented Reality/Virtual Reality (AR/VR), among other areas. BCI which can decode and recognize neural signals involved in speech and handwriting has the potential to greatly assist individuals with severe motor impairments in their communication and interaction needs. Innovative and cutting-edge advancements in this field have the potential to develop a highly accessible and interactive communication platform for these people. The purpose of this review paper is to analyze the existing research on handwriting and speech recognition from neural signals. So that the new researchers who are interested in this field can gain thorough knowledge in this research area. The current research on neural signal-based recognition of handwriting and speech has been categorized into two main types: invasive and non-invasive studies. We have examined the latest papers on converting speech-activity-based neural signals and handwriting-activity-based neural signals into text data. The methods of extracting data from the brain have also been discussed in this review. Additionally, this review includes a brief summary of the datasets, preprocessing techniques, and methods used in these studies, which were published between 2014 and 2022. This review aims to provide a comprehensive summary of the methodologies used in the current literature on neural signal-based recognition of handwriting and speech. In essence, this article is intended to serve as a valuable resource for future researchers who wish to investigate neural signal-based machine-learning methods in their work.

Linguistic representation of vowels in speech imagery EEG

Speech imagery recognition from electroencephalograms (EEGs) could potentially become a strong contender among non-invasive brain-computer interfaces (BCIs). In this report, first we extract language representations as the difference of line-spectra of phones by statistically analyzing many EEG signals from the Broca area. Then we extract vowels by using iterative search from hand-labeled short-syllable data. The iterative search process consists of principal component analysis (PCA) that visualizes linguistic representation of vowels through eigen-vectors φ(m), and subspace method (SM) that searches an optimum line-spectrum for redesigning φ(m). The extracted linguistic representation of Japanese vowels /i/ /e/ /a/ /o/ /u/ shows 2 distinguished spectral peaks (P1, P2) in the upper frequency range. The 5 vowels are aligned on the P1-P2 chart. A 5-vowel recognition experiment using a data set of 5 subjects and a convolutional neural network (CNN) classifier gave a mean accuracy rate of 72.6%.

Closed-Loop Adaptive Deep Brain Stimulation in Parkinson's Disease: Procedures to Achieve It and Future Perspectives

Parkinson's disease (PD) is a neurodegenerative disease with a heavy burden on patients, families, and society. Deep brain stimulation (DBS) can improve the symptoms of PD patients for whom medication is insufficient. However, current open-loop uninterrupted conventional DBS (cDBS) has inherent limitations, such as adverse effects, rapid battery consumption, and a need for frequent parameter adjustment. To overcome these shortcomings, adaptive DBS (aDBS) was proposed to provide responsive optimized stimulation for PD. This topic has attracted scientific interest, and a growing body of preclinical and clinical evidence has shown its benefits. However, both achievements and challenges have emerged in this novel field. To date, only limited reviews comprehensively analyzed the full framework and procedures for aDBS implementation. Herein, we review current preclinical and clinical data on aDBS for PD to discuss the full procedures for its achievement and to provide future perspectives on this treatment.

Intelligent speech technologies for transcription, disease diagnosis, and medical equipment interactive control in smart hospitals: A review

The growing and aging of the world population have driven the shortage of medical resources in recent years, especially during the COVID-19 pandemic. Fortunately, the rapid development of robotics and artificial intelligence technologies help to adapt to the challenges in the healthcare field. Among them, intelligent speech technology (IST) has served doctors and patients to improve the efficiency of medical behavior and alleviate the medical burden. However, problems like noise interference in complex medical scenarios and pronunciation differences between patients and healthy people hamper the broad application of IST in hospitals. In recent years, technologies such as machine learning have developed rapidly in intelligent speech recognition, which is expected to solve these problems. This paper first introduces IST's procedure and system architecture and analyzes its application in medical scenarios. Secondly, we review existing IST applications in smart hospitals in detail, including electronic medical documentation, disease diagnosis and evaluation, and human-medical equipment interaction. In addition, we elaborate on an application case of IST in the early recognition, diagnosis, rehabilitation training, evaluation, and daily care of stroke patients. Finally, we discuss IST's limitations, challenges, and future directions in the medical field. Furthermore, we propose a novel medical voice analysis system architecture that employs active hardware, active software, and human-computer interaction to realize intelligent and evolvable speech recognition. This comprehensive review and the proposed architecture offer directions for future studies on IST and its applications in smart hospitals.

Dataset of Speech Production in intracranial.Electroencephalography

Speech production is an intricate process involving a large number of muscles and cognitive processes. The neural processes underlying speech production are not completely understood. As speech is a uniquely human ability, it can not be investigated in animal models. High-fidelity human data can only be obtained in clinical settings and is therefore not easily available to all researchers. Here, we provide a dataset of 10 participants reading out individual words while we measured intracranial EEG from a total of 1103 electrodes. The data, with its high temporal resolution and coverage of a large variety of cortical and sub-cortical brain regions, can help in understanding the speech production process better. Simultaneously, the data can be used to test speech decoding and synthesis approaches from neural data to develop speech Brain-Computer Interfaces and speech neuroprostheses.

Measurement(s)	Brain activity
Technology Type(s)	Stereotactic electroencephalography
Sample Characteristic - Organism	Homo sapiens
Sample Characteristic - Environment	Epilepsy monitoring center
Sample Characteristic - Location	The Netherlands

Decoding lexical tones and vowels in imagined tonal monosyllables using fNIRS signals

Objective.Speech is a common way of communication. Decoding verbal intent could provide a naturalistic communication way for people with severe motor disabilities. Active brain computer interaction (BCI) speller is one of the most commonly used speech BCIs. To reduce the spelling time of Chinese words, identifying vowels and tones that are embedded in imagined Chinese words is essential. Functional near-infrared spectroscopy (fNIRS) has been widely used in BCI because it is portable, non-invasive, safe, low cost, and has a relatively high spatial resolution.Approach.In this study, an active BCI speller based on fNIRS is presented by covertly rehearsing tonal monosyllables with vowels (i.e. /a/, /i/, /o/, and /u/) and four lexical tones in Mandarin Chinese (i.e. tones 1, 2, 3, and 4) for 10 s.Main results.fNIRS results showed significant differences in the right superior temporal gyrus between imagined vowels with tone 2/3/4 and those with tone 1 (i.e. more activations and stronger connections to other brain regions for imagined vowels with tones 2/3/4 than for those with tone 1). Speech-related areas for tone imagery (i.e. the right hemisphere) provided majority of information for identifying tones, while the left hemisphere had advantages in vowel identification. Having decoded both vowels and tones during the post-stimulus 15 s period, the average classification accuracies exceeded 40% and 70% in multiclass (i.e. four classes) and binary settings, respectively. To spell words more quickly, the time window size for decoding was reduced from 15 s to 2.5 s while the classification accuracies were not significantly reduced.Significance.For the first time, this work demonstrated the possibility of discriminating lexical tones and vowels in imagined tonal syllables simultaneously. In addition, the reduced time window for decoding indicated that the spelling time of Chinese words could be significantly reduced in the fNIRS-based BCIs.

The Role of Artificial Intelligence in Decoding Speech from EEG Signals: A Scoping Review

Background: Brain traumas, mental disorders, and vocal abuse can result in permanent or temporary speech impairment, significantly impairing one's quality of life and occasionally resulting in social isolation. Brain-computer interfaces (BCI) can support people who have issues with their speech or who have been paralyzed to communicate with their surroundings via brain signals. Therefore, EEG signal-based BCI has received significant attention in the last two decades for multiple reasons: (i) clinical research has capitulated detailed knowledge of EEG signals, (ii) inexpensive EEG devices, and (iii) its application in medical and social fields. Objective: This study explores the existing literature and summarizes EEG data acquisition, feature extraction, and artificial intelligence (AI) techniques for decoding speech from brain signals. Method: We followed the PRISMA-ScR guidelines to conduct this scoping review. We searched six electronic databases: PubMed, IEEE Xplore, the ACM Digital Library, Scopus, arXiv, and Google Scholar. We carefully selected search terms based on target intervention (i.e., imagined speech and AI) and target data (EEG signals), and some of the search terms were derived from previous reviews. The study selection process was carried out in three phases: study identification, study selection, and data extraction. Two reviewers independently carried out study selection and data extraction. A narrative approach was adopted to synthesize the extracted data. Results: A total of 263 studies were evaluated; however, 34 met the eligibility criteria for inclusion in this review. We found 64-electrode EEG signal devices to be the most widely used in the included studies. The most common signal normalization and feature extractions in the included studies were the bandpass filter and wavelet-based feature extraction. We categorized the studies based on AI techniques, such as machine learning and deep learning. The most prominent ML algorithm was a support vector machine, and the DL algorithm was a convolutional neural network. Conclusions: EEG signal-based BCI is a viable technology that can enable people with severe or temporal voice impairment to communicate to the world directly from their brain. However, the development of BCI technology is still in its infancy.

Harnessing the Power of Artificial Intelligence in Otolaryngology and the Communication Sciences

Use of artificial intelligence (AI) is a burgeoning field in otolaryngology and the communication sciences. A virtual symposium on the topic was convened from Duke University on October 26, 2020, and was attended by more than 170 participants worldwide. This review presents summaries of all but one of the talks presented during the symposium; recordings of all the talks, along with the discussions for the talks, are available at https://www.youtube.com/watch?v=ktfewrXvEFg and https://www.youtube.com/watch?v=-gQ5qX2v3rg . Each of the summaries is about 2500 words in length and each summary includes two figures. This level of detail far exceeds the brief summaries presented in traditional reviews and thus provides a more-informed glimpse into the power and diversity of current AI applications in otolaryngology and the communication sciences and how to harness that power for future applications.

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

Currently, the most used method to measure brain activity under a non-invasive procedure is the electroencephalogram (EEG). This is because of its high temporal resolution, ease of use, and safety. These signals can be used under a Brain Computer Interface (BCI) framework, which can be implemented to provide a new communication channel to people that are unable to speak due to motor disabilities or other neurological diseases. Nevertheless, EEG-based BCI systems have presented challenges to be implemented in real life situations for imagined speech recognition due to the difficulty to interpret EEG signals because of their low signal-to-noise ratio (SNR). As consequence, in order to help the researcher make a wise decision when approaching this problem, we offer a review article that sums the main findings of the most relevant studies on this subject since 2009. This review focuses mainly on the pre-processing, feature extraction, and classification techniques used by several authors, as well as the target vocabulary. Furthermore, we propose ideas that may be useful for future work in order to achieve a practical application of EEG-based BCI systems toward imagined speech decoding.

Predictive accuracy of CNN for cortical oscillatory activity in an acute rat model of parkinsonism

In neurological and neuropsychiatric disorders neuronal oscillatory activity between basal ganglia and cortical circuits are altered, which may be useful as biomarker for adaptive deep brain stimulation. We investigated whether changes in the spectral power of oscillatory activity in the motor cortex (MCtx) and the sensorimotor cortex (SMCtx) of rats after injection of the dopamine (DA) receptor antagonist haloperidol (HALO) would be similar to those observed in Parkinson disease. Thereafter, we tested whether a convolutional neural network (CNN) model would identify brain signal alterations in this acute model of parkinsonism. A sixteen channel surface micro-electrocorticogram (ECoG) recording array was placed under the dura above the MCtx and SMCtx areas of one hemisphere under general anaesthesia in rats. Seven days after surgery, micro ECoG was recorded in individual free moving rats in three conditions: (1) basal activity, (2) after injection of HALO (0.5 mg/kg), and (3) with additional injection of apomorphine (APO) (1 mg/kg). Furthermore, a CNN-based classification consisting of 23,530 parameters was applied on the raw data. HALO injection decreased oscillatory theta band activity (4-8 Hz) and enhanced beta (12-30 Hz) and gamma (30-100 Hz) in MCtx and SMCtx, which was compensated after APO injection (P ¡ 0.001). Evaluation of classification performance of the CNN model provided accuracy of 92%, sensitivity of 90% and specificity of 93% on one-dimensional signals. The CNN proposed model requires a minimum of sensory hardware and may be integrated into future research on therapeutic devices for Parkinson disease, such as adaptive closed loop stimulation, thus contributing to more efficient way of treatment.

Speech Synthesis from Stereotactic EEG using an Electrode Shaft Dependent Multi-Input Convolutional Neural Network Approach

Neurological disorders can lead to significant impairments in speech communication and, in severe cases, cause the complete loss of the ability to speak. Brain-Computer Interfaces have shown promise as an alternative communication modality by directly transforming neural activity of speech processes into a textual or audible representations. Previous studies investigating such speech neuroprostheses relied on electrocorticography (ECoG) or microelectrode arrays that acquire neural signals from superficial areas on the cortex. While both measurement methods have demonstrated successful speech decoding, they do not capture activity from deeper brain structures and this activity has therefore not been harnessed for speech-related BCIs. In this study, we bridge this gap by adapting a previously presented decoding pipeline for speech synthesis based on ECoG signals to implanted depth electrodes (sEEG). For this purpose, we propose a multi-input convolutional neural network that extracts speech-related activity separately for each electrode shaft and estimates spectral coefficients to reconstruct an audible waveform. We evaluate our approach on open-loop data from 5 patients who conducted a recitation task of Dutch utterances. We achieve correlations of up to 0.80 between original and reconstructed speech spectrograms, which are significantly above chance level for all patients (p < 0.001). Our results indicate that sEEG can yield similar speech decoding performance to prior ECoG studies and is a promising modality for speech BCIs.

Imagined speech increases the hemodynamic response and functional connectivity of the dorsal motor cortex

Objective. Decoding imagined speech from brain signals could provide a more natural, user-friendly way for developing the next generation of the brain-computer interface (BCI). With the advantages of non-invasive, portable, relatively high spatial resolution and insensitivity to motion artifacts, the functional near-infrared spectroscopy (fNIRS) shows great potential for developing the non-invasive speech BCI. However, there is a lack of fNIRS evidence in uncovering the neural mechanism of imagined speech. Our goal is to investigate the specific brain regions and the corresponding cortico-cortical functional connectivity features during imagined speech with fNIRS.Approach. fNIRS signals were recorded from 13 subjects' bilateral motor and prefrontal cortex during overtly and covertly repeating words. Cortical activation was determined through the mean oxygen-hemoglobin concentration changes, and functional connectivity was calculated by Pearson's correlation coefficient.Main results. (a) The bilateral dorsal motor cortex was significantly activated during the covert speech, whereas the bilateral ventral motor cortex was significantly activated during the overt speech. (b) As a subregion of the motor cortex, sensorimotor cortex (SMC) showed a dominant dorsal response to covert speech condition, whereas a dominant ventral response to overt speech condition. (c) Broca's area was deactivated during the covert speech but activated during the overt speech. (d) Compared to overt speech, dorsal SMC(dSMC)-related functional connections were enhanced during the covert speech.Significance. We provide fNIRS evidence for the involvement of dSMC in speech imagery. dSMC is the speech imagery network's key hub and is probably involved in the sensorimotor information processing during the covert speech. This study could inspire the BCI community to focus on the potential contribution of dSMC during speech imagery.

Local field potentials in a pre-motor region predict learned vocal sequences

Neuronal activity within the premotor region HVC is tightly synchronized to, and crucial for, the articulate production of learned song in birds. Characterizations of this neural activity detail patterns of sequential bursting in small, carefully identified subsets of neurons in the HVC population. The dynamics of HVC are well described by these characterizations, but have not been verified beyond this scale of measurement. There is a rich history of using local field potentials (LFP) to extract information about behavior that extends beyond the contribution of individual cells. These signals have the advantage of being stable over longer periods of time, and they have been used to study and decode human speech and other complex motor behaviors. Here we characterize LFP signals presumptively from the HVC of freely behaving male zebra finches during song production to determine if population activity may yield similar insights into the mechanisms underlying complex motor-vocal behavior. Following an initial observation that structured changes in the LFP were distinct to all vocalizations during song, we show that it is possible to extract time-varying features from multiple frequency bands to decode the identity of specific vocalization elements (syllables) and to predict their temporal onsets within the motif. This demonstrates the utility of LFP for studying vocal behavior in songbirds. Surprisingly, the time frequency structure of HVC LFP is qualitatively similar to well-established oscillations found in both human and non-human mammalian motor areas. This physiological similarity, despite distinct anatomical structures, may give insight into common computational principles for learning and/or generating complex motor-vocal behaviors.

Vocalizations, such as speech and song, are a motor process that requires the coordination of numerous muscle groups receiving instructions from specific brain regions. In songbirds, HVC is a premotor brain region required for singing; it is populated by a set of neurons that fire sparsely during song. How HVC enables song generation is not well understood. Here we describe network activity presumptively from HVC that precedes the initiation of each vocal element during singing. This network activity can be used to predict both the identity of each vocal element (syllable) and when it will occur during song. In addition, this network activity is similar to activity that has been documented in human, non-human primate, and mammalian premotor regions tied to muscle movements. These similarities add to a growing body of literature that finds parallels between songbirds and humans in respect to the motor control of vocal organs. Furthermore, given the similarities of the songbird and human motor-vocal systems, these results suggest that the songbird model could be leveraged to accelerate the development of clinically translatable speech prosthesis.

Decoding Covert Speech From EEG-A Comprehensive Review

Over the past decade, many researchers have come up with different implementations of systems for decoding covert or imagined speech from EEG (electroencephalogram). They differ from each other in several aspects, from data acquisition to machine learning algorithms, due to which, a comparison between different implementations is often difficult. This review article puts together all the relevant works published in the last decade on decoding imagined speech from EEG into a single framework. Every important aspect of designing such a system, such as selection of words to be imagined, number of electrodes to be recorded, temporal and spatial filtering, feature extraction and classifier are reviewed. This helps a researcher to compare the relative merits and demerits of the different approaches and choose the one that is most optimal. Speech being the most natural form of communication which human beings acquire even without formal education, imagined speech is an ideal choice of prompt for evoking brain activity patterns for a BCI (brain-computer interface) system, although the research on developing real-time (online) speech imagery based BCI systems is still in its infancy. Covert speech based BCI can help people with disabilities to improve their quality of life. It can also be used for covert communication in environments that do not support vocal communication. This paper also discusses some future directions, which will aid the deployment of speech imagery based BCI for practical applications, rather than only for laboratory experiments.

Adaptive latent state modeling of brain network dynamics with real-time learning rate optimization

Objective. Dynamic latent state models are widely used to characterize the dynamics of brain network activity for various neural signal types. To date, dynamic latent state models have largely been developed for stationary brain network dynamics. However, brain network dynamics can be non-stationary for example due to learning, plasticity or recording instability. To enable modeling these non-stationarities, two problems need to be resolved. First, novel methods should be developed that can adaptively update the parameters of latent state models, which is difficult due to the state being latent. Second, new methods are needed to optimize the adaptation learning rate, which specifies how fast new neural observations update the model parameters and can significantly influence adaptation accuracy.Approach. We develop a Rate Optimized-adaptive Linear State-Space Modeling (RO-adaptive LSSM) algorithm that solves these two problems. First, to enable adaptation, we derive a computation- and memory-efficient adaptive LSSM fitting algorithm that updates the LSSM parameters recursively and in real time in the presence of the latent state. Second, we develop a real-time learning rate optimization algorithm. We use comprehensive simulations of a broad range of non-stationary brain network dynamics to validate both algorithms, which together constitute the RO-adaptive LSSM.Main results. We show that the adaptive LSSM fitting algorithm can accurately track the broad simulated non-stationary brain network dynamics. We also find that the learning rate significantly affects the LSSM fitting accuracy. Finally, we show that the real-time learning rate optimization algorithm can run in parallel with the adaptive LSSM fitting algorithm. Doing so, the combined RO-adaptive LSSM algorithm rapidly converges to the optimal learning rate and accurately tracks non-stationarities.Significance. These algorithms can be used to study time-varying neural dynamics underlying various brain functions and enhance future neurotechnologies such as brain-machine interfaces and closed-loop brain stimulation systems.

Speech Technology for Healthcare: Opportunities, Challenges, and State of the Art

Speech technology is not appropriately explored even though modern advances in speech technology-especially those driven by deep learning (DL) technology-offer unprecedented opportunities for transforming the healthcare industry. In this paper, we have focused on the enormous potential of speech technology for revolutionising the healthcare domain. More specifically, we review the state-of-the-art approaches in automatic speech recognition (ASR), speech synthesis or text to speech (TTS), and health detection and monitoring using speech signals. We also present a comprehensive overview of various challenges hindering the growth of speech-based services in healthcare. To make speech-based healthcare solutions more prevalent, we discuss open issues and suggest some possible research directions aimed at fully leveraging the advantages of other technologies for making speech-based healthcare solutions more effective.

Deep learning approaches for neural decoding across architectures and recording modalities

Decoding behavior, perception or cognitive state directly from neural signals is critical for brain-computer interface research and an important tool for systems neuroscience. In the last decade, deep learning has become the state-of-the-art method in many machine learning tasks ranging from speech recognition to image segmentation. The success of deep networks in other domains has led to a new wave of applications in neuroscience. In this article, we review deep learning approaches to neural decoding. We describe the architectures used for extracting useful features from neural recording modalities ranging from spikes to functional magnetic resonance imaging. Furthermore, we explore how deep learning has been leveraged to predict common outputs including movement, speech and vision, with a focus on how pretrained deep networks can be incorporated as priors for complex decoding targets like acoustic speech or images. Deep learning has been shown to be a useful tool for improving the accuracy and flexibility of neural decoding across a wide range of tasks, and we point out areas for future scientific development.

Decoding spoken English from intracortical electrode arrays in dorsal precentral gyrus

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.

Modeling behaviorally relevant neural dynamics enabled by preferential subspace identification

Neural activity exhibits complex dynamics related to various brain functions, internal states and behaviors. Understanding how neural dynamics explain specific measured behaviors requires dissociating behaviorally relevant and irrelevant dynamics, which is not achieved with current neural dynamic models as they are learned without considering behavior. We develop preferential subspace identification (PSID), which is an algorithm that models neural activity while dissociating and prioritizing its behaviorally relevant dynamics. Modeling data in two monkeys performing three-dimensional reach and grasp tasks, PSID revealed that the behaviorally relevant dynamics are significantly lower-dimensional than otherwise implied. Moreover, PSID discovered distinct rotational dynamics that were more predictive of behavior. Furthermore, PSID more accurately learned behaviorally relevant dynamics for each joint and recording channel. Finally, modeling data in two monkeys performing saccades demonstrated the generalization of PSID across behaviors, brain regions and neural signal types. PSID provides a general new tool to reveal behaviorally relevant neural dynamics that can otherwise go unnoticed.

The Potential of Stereotactic-EEG for Brain-Computer Interfaces: Current Progress and Future Directions

Stereotactic electroencephalogaphy (sEEG) utilizes localized, penetrating depth electrodes to measure electrophysiological brain activity. It is most commonly used in the identification of epileptogenic zones in cases of refractory epilepsy. The implanted electrodes generally provide a sparse sampling of a unique set of brain regions including deeper brain structures such as hippocampus, amygdala and insula that cannot be captured by superficial measurement modalities such as electrocorticography (ECoG). Despite the overlapping clinical application and recent progress in decoding of ECoG for Brain-Computer Interfaces (BCIs), sEEG has thus far received comparatively little attention for BCI decoding. Additionally, the success of the related deep-brain stimulation (DBS) implants bodes well for the potential for chronic sEEG applications. This article provides an overview of sEEG technology, BCI-related research, and prospective future directions of sEEG for long-term BCI applications.

Electrocorticogram (ECoG) Is Highly Informative in Primate Visual Cortex

Neural signals recorded at different scales contain information about environment and behavior and have been used to control Brain Machine Interfaces with varying degrees of success. However, a direct comparison of their efficacy has not been possible due to different recording setups, tasks, species, etc. To address this, we implanted customized arrays having both microelectrodes and electrocorticogram (ECoG) electrodes in the primary visual cortex of 2 female macaque monkeys, and also recorded electroencephalogram (EEG), while they viewed a variety of naturalistic images and parametric gratings. Surprisingly, ECoG had higher information and decodability than all other signals. Combining a few ECoG electrodes allowed more accurate decoding than combining a much larger number of microelectrodes. Control analyses showed that higher decoding accuracy of ECoG compared with local field potential was not because of differences in low-level visual features captured by them but instead because of larger spatial summation of the ECoG. Information was high in the 30-80 Hz range and at lower frequencies. Information in different frequencies and scales was nonredundant. These results have strong implications for Brain Machine Interface applications and for study of population representation of visual stimuli.SIGNIFICANCE STATEMENT Electrophysiological signals captured across scales by different recording electrodes are regularly used for Brain Machine Interfaces, but the information content varies due to electrode size and location. A systematic comparison of their efficiency for Brain Machine Interfaces is important but technically challenging. Here, we recorded simultaneous signals across four scales: spikes, local field potential, electrocorticogram (ECoG), and EEG, and compared their information and decoding accuracy for a large variety of naturalistic stimuli. We found that ECoGs were highly informative and outperformed other signals in information content and decoding accuracy.

Towards Restoration of Articulatory Movements: Functional Electrical Stimulation of Orofacial Muscles

Millions of individuals suffer from impairments that significantly disrupt or completely eliminate their ability to speak. An ideal intervention would restore one's natural ability to physically produce speech. Recent progress has been made in decoding speech-related brain activity to generate synthesized speech. Our vision is to extend these recent advances toward the goal of restoring physical speech production using decoded speech-related brain activity to modulate the electrical stimulation of the orofacial musculature involved in speech. In this pilot study we take a step toward this vision by investigating the feasibility of stimulating orofacial muscles during vocalization in order to alter acoustic production. The results of our study provide necessary foundation for eventual orofacial stimulation controlled directly from decoded speech-related brain activity.

Neural ensemble dynamics in dorsal motor cortex during speech in people with paralysis

Speaking is a sensorimotor behavior whose neural basis is difficult to study with single neuron resolution due to the scarcity of human intracortical measurements. We used electrode arrays to record from the motor cortex 'hand knob' in two people with tetraplegia, an area not previously implicated in speech. Neurons modulated during speaking and during non-speaking movements of the tongue, lips, and jaw. This challenges whether the conventional model of a 'motor homunculus' division by major body regions extends to the single-neuron scale. Spoken words and syllables could be decoded from single trials, demonstrating the potential of intracortical recordings for brain-computer interfaces to restore speech. Two neural population dynamics features previously reported for arm movements were also present during speaking: a component that was mostly invariant across initiating different words, followed by rotatory dynamics during speaking. This suggests that common neural dynamical motifs may underlie movement of arm and speech articulators.

Workshops of the Seventh International Brain-Computer Interface Meeting: Not Getting Lost in Translation

The Seventh International Brain-Computer Interface (BCI) Meeting was held May 21-25th, 2018 at the Asilomar Conference Grounds, Pacific Grove, California, United States. The interactive nature of this conference was embodied by 25 workshops covering topics in BCI (also called brain-machine interface) research. Workshops covered foundational topics such as hardware development and signal analysis algorithms, new and imaginative topics such as BCI for virtual reality and multi-brain BCIs, and translational topics such as clinical applications and ethical assumptions of BCI development. BCI research is expanding in the diversity of applications and populations for whom those applications are being developed. BCI applications are moving toward clinical readiness as researchers struggle with the practical considerations to make sure that BCI translational efforts will be successful. This paper summarizes each workshop, providing an overview of the topic of discussion, references for additional information, and identifying future issues for research and development that resulted from the interactions and discussion at the workshop.

Determining the Optimal Number of MEG Trials: A Machine Learning and Speech Decoding Perspective

Advancing the knowledge about neural speech mechanisms is critical for developing next-generation, faster brain computer interface to assist in speech communication for the patients with severe neurological conditions (e.g., locked-in syndrome). Among current neuroimaging techniques, Magnetoencephalography (MEG) provides direct representation for the large-scale neural dynamics of underlying cognitive processes based on its optimal spatiotemporal resolution. However, the MEG measured neural signals are smaller in magnitude compared to the background noise and hence, MEG usually suffers from a low signal-to-noise ratio (SNR) at the single-trial level. To overcome this limitation, it is common to record many trials of the same event-task and use the time-locked average signal for analysis, which can be very time consuming. In this study, we investigated the effect of the number of MEG recording trials required for speech decoding using a machine learning algorithm. We used a wavelet filter for generating the denoised neural features to train an Artificial Neural Network (ANN) for speech decoding. We found that wavelet based denoising increased the SNR of the neural signal prior to analysis and facilitated accurate speech decoding performance using as few as 40 single-trials. This study may open up the possibility of limiting MEG trials for other task evoked studies as well.

Generating Natural, Intelligible Speech From Brain Activity in Motor, Premotor, and Inferior Frontal Cortices

Neural interfaces that directly produce intelligible speech from brain activity would allow people with severe impairment from neurological disorders to communicate more naturally. Here, we record neural population activity in motor, premotor and inferior frontal cortices during speech production using electrocorticography (ECoG) and show that ECoG signals alone can be used to generate intelligible speech output that can preserve conversational cues. To produce speech directly from neural data, we adapted a method from the field of speech synthesis called unit selection, in which units of speech are concatenated to form audible output. In our approach, which we call Brain-To-Speech, we chose subsequent units of speech based on the measured ECoG activity to generate audio waveforms directly from the neural recordings. Brain-To-Speech employed the user's own voice to generate speech that sounded very natural and included features such as prosody and accentuation. By investigating the brain areas involved in speech production separately, we found that speech motor cortex provided more information for the reconstruction process than the other cortical areas.

Decoding Speech from Intracortical Multielectrode Arrays in Dorsal "Arm/Hand Areas" of Human Motor Cortex

Neural prostheses are being developed to restore speech to people with neurological injury or disease. A key design consideration is where and how to access neural correlates of intended speech. Most prior work has examined cortical field potentials at a coarse resolution using electroencephalography (EEG) or medium resolution using electrocorticography (ECoG). The few studies of speech with single-neuron resolution recorded from ventral areas known to be part of the speech network. Here, we recorded from two 96- electrode arrays chronically implanted into the 'hand knob' area of motor cortex while a person with tetraplegia spoke. Despite being located in an area previously demonstrated to modulate during attempted arm movements, many electrodes' neuronal firing rates responded to speech production. In offline analyses, we could classify which of 9 phonemes (plus silence) was spoken with 81% single-trial accuracy using a combination of spike rate and local field potential (LFP) power. This suggests that high-fidelity speech prostheses may be possible using large-scale intracortical recordings in motor cortical areas involved in controlling speech articulators.

Human motor decoding from neural signals: a review

Many people suffer from movement disability due to amputation or neurological diseases. Fortunately, with modern neurotechnology now it is possible to intercept motor control signals at various points along the neural transduction pathway and use that to drive external devices for communication or control. Here we will review the latest developments in human motor decoding. We reviewed the various strategies to decode motor intention from human and their respective advantages and challenges. Neural control signals can be intercepted at various points in the neural signal transduction pathway, including the brain (electroencephalography, electrocorticography, intracortical recordings), the nerves (peripheral nerve recordings) and the muscles (electromyography). We systematically discussed the sites of signal acquisition, available neural features, signal processing techniques and decoding algorithms in each of these potential interception points. Examples of applications and the current state-of-the-art performance were also reviewed. Although great strides have been made in human motor decoding, we are still far away from achieving naturalistic and dexterous control like our native limbs. Concerted efforts from material scientists, electrical engineers, and healthcare professionals are needed to further advance the field and make the technology widely available in clinical use.

Speech synthesis from ECoG using densely connected 3D convolutional neural networks

OBJECTIVE

Direct synthesis of speech from neural signals could provide a fast and natural way of communication to people with neurological diseases. Invasively-measured brain activity (electrocorticography; ECoG) supplies the necessary temporal and spatial resolution to decode fast and complex processes such as speech production. A number of impressive advances in speech decoding using neural signals have been achieved in recent years, but the complex dynamics are still not fully understood. However, it is unlikely that simple linear models can capture the relation between neural activity and continuous spoken speech.

APPROACH

Here we show that deep neural networks can be used to map ECoG from speech production areas onto an intermediate representation of speech (logMel spectrogram). The proposed method uses a densely connected convolutional neural network topology which is well-suited to work with the small amount of data available from each participant.

MAIN RESULTS

In a study with six participants, we achieved correlations up to r  =  0.69 between the reconstructed and original logMel spectrograms. We transfered our prediction back into an audible waveform by applying a Wavenet vocoder. The vocoder was conditioned on logMel features that harnessed a much larger, pre-existing data corpus to provide the most natural acoustic output.

SIGNIFICANCE

To the best of our knowledge, this is the first time that high-quality speech has been reconstructed from neural recordings during speech production using deep neural networks.

Towards reconstructing intelligible speech from the human auditory cortex

Auditory stimulus reconstruction is a technique that finds the best approximation of the acoustic stimulus from the population of evoked neural activity. Reconstructing speech from the human auditory cortex creates the possibility of a speech neuroprosthetic to establish a direct communication with the brain and has been shown to be possible in both overt and covert conditions. However, the low quality of the reconstructed speech has severely limited the utility of this method for brain-computer interface (BCI) applications. To advance the state-of-the-art in speech neuroprosthesis, we combined the recent advances in deep learning with the latest innovations in speech synthesis technologies to reconstruct closed-set intelligible speech from the human auditory cortex. We investigated the dependence of reconstruction accuracy on linear and nonlinear (deep neural network) regression methods and the acoustic representation that is used as the target of reconstruction, including auditory spectrogram and speech synthesis parameters. In addition, we compared the reconstruction accuracy from low and high neural frequency ranges. Our results show that a deep neural network model that directly estimates the parameters of a speech synthesizer from all neural frequencies achieves the highest subjective and objective scores on a digit recognition task, improving the intelligibility by 65% over the baseline method which used linear regression to reconstruct the auditory spectrogram. These results demonstrate the efficacy of deep learning and speech synthesis algorithms for designing the next generation of speech BCI systems, which not only can restore communications for paralyzed patients but also have the potential to transform human-computer interaction technologies.

An update on adaptive deep brain stimulation in Parkinson's disease

Advancing conventional open‐loop DBS as a therapy for PD is crucial for overcoming important issues such as the delicate balance between beneficial and adverse effects and limited battery longevity that are currently associated with treatment. Closed‐loop or adaptive DBS aims to overcome these limitations by real‐time adjustment of stimulation parameters based on continuous feedback input signals that are representative of the patient's clinical state. The focus of this update is to discuss the most recent developments regarding potential input signals and possible stimulation parameter modulation for adaptive DBS in PD. Potential input signals for adaptive DBS include basal ganglia local field potentials, cortical recordings (electrocorticography), wearable sensors, and eHealth and mHealth devices. Furthermore, adaptive DBS can be applied with different approaches of stimulation parameter modulation, the feasibility of which can be adapted depending on specific PD phenotypes. Implementation of technological developments like machine learning show potential in the design of such approaches; however, energy consumption deserves further attention. Furthermore, we discuss future considerations regarding the clinical implementation of adaptive DBS in PD. © 2018 The Authors. Movement Disorders published by Wiley Periodicals, Inc. on behalf of International Parkinson and Movement Disorder Society.

Speech Recognition via fNIRS Based Brain Signals

In this paper, we present the first evidence that perceived speech can be identified from the listeners' brain signals measured via functional-near infrared spectroscopy (fNIRS)—a non-invasive, portable, and wearable neuroimaging technique suitable for ecologically valid settings. In this study, participants listened audio clips containing English stories while prefrontal and parietal cortices were monitored with fNIRS. Machine learning was applied to train predictive models using fNIRS data from a subject pool to predict which part of a story was listened by a new subject not in the pool based on the brain's hemodynamic response as measured by fNIRS. fNIRS signals can vary considerably from subject to subject due to the different head size, head shape, and spatial locations of brain functional regions. To overcome this difficulty, a generalized canonical correlation analysis (GCCA) was adopted to extract latent variables that are shared among the listeners before applying principal component analysis (PCA) for dimension reduction and applying logistic regression for classification. A 74.7% average accuracy has been achieved for differentiating between two 50 s. long story segments and a 43.6% average accuracy has been achieved for differentiating four 25 s. long story segments. These results suggest the potential of an fNIRS based-approach for building a speech decoding brain-computer-interface for developing a new type of neural prosthetic system.

Vowel Imagery Decoding toward Silent Speech BCI Using Extreme Learning Machine with Electroencephalogram

The purpose of this study is to classify EEG data on imagined speech in a single trial. We recorded EEG data while five subjects imagined different vowels, /a/, /e/, /i/, /o/, and /u/. We divided each single trial dataset into thirty segments and extracted features (mean, variance, standard deviation, and skewness) from all segments. To reduce the dimension of the feature vector, we applied a feature selection algorithm based on the sparse regression model. These features were classified using a support vector machine with a radial basis function kernel, an extreme learning machine, and two variants of an extreme learning machine with different kernels. Because each single trial consisted of thirty segments, our algorithm decided the label of the single trial by selecting the most frequent output among the outputs of the thirty segments. As a result, we observed that the extreme learning machine and its variants achieved better classification rates than the support vector machine with a radial basis function kernel and linear discrimination analysis. Thus, our results suggested that EEG responses to imagined speech could be successfully classified in a single trial using an extreme learning machine with a radial basis function and linear kernel. This study with classification of imagined speech might contribute to the development of silent speech BCI systems.