Model-Based and Model-Free Analyses of the Neural Correlates of Tongue Movements

Palatal Electrotactile Display Outperforms Visual Display in Tongue Motor learning

Incomplete tongue motor control is a common yet challenging issue among individuals with neurotraumas and neurological disorders. In development of the training protocols, multiple sensory modalities including visual, auditory, and tactile feedback have been employed. However, the effectiveness of each sensory modality in tongue motor learning is still in question. The object of this study was to test the effectiveness of visual and electrotactile assistance on tongue motor learning, respectively. Eight healthy subjects performed the tongue pointing task, in which they were visually instructed to touch the target on the palate by their tongue tip as accurately as possible. Each subject wore a custom-made dental retainer with 12 electrodes distributed over the palatal area. For visual training, 3×4 LED array on the computer screen, corresponding to the electrode layout, was turned on with different colors according to the tongue contact. For electrotactile training, electrical stimulation was applied to the tongue with frequencies depending on the distance between the tongue contact and the target, along with a small protrusion on the retainer as an indicator of the target. One baseline session, one training session, and three post-training sessions were conducted over four-day duration. Experimental result showed that the error was decreased after both visual and electrotactile trainings, from 3.56±0.11 (Mean±STE) to 1.27±0.16, and from 3.97±0.11 to 0.53±0.19, respectively. The result also showed that electrotactile training leads to stronger retention than visual training, as the improvement was retained as 62.68±1.81% after electrotactile training and 36.59±2.24% after visual training, at 3-day post training.

Functional MRI of Native and Non-native Speech Sound Production in Sequential German-English Bilinguals

Bilingualism and multilingualism are highly prevalent. Non-invasive brain imaging has been used to study the neural correlates of native and non-native speech and language production, mainly on the lexical and syntactic level. Here, we acquired continuous fast event-related FMRI during visually cued overt production of exclusively German and English vowels and syllables. We analyzed data from 13 university students, native speakers of German and sequential English bilinguals. The production of non-native English sounds was associated with increased activity of the left primary sensorimotor cortex, bilateral cerebellar hemispheres (lobule VI), left inferior frontal gyrus, and left anterior insula compared to native German sounds. The contrast German > English sounds was not statistically significant. Our results emphasize that the production of non-native speech requires additional neural resources already on a basic phonological level in sequential bilinguals.

NeuroVAD: Real-Time Voice Activity Detection from Non-Invasive Neuromagnetic Signals

Neural speech decoding-driven brain-computer interface (BCI) or speech-BCI is a novel paradigm for exploring communication restoration for locked-in (fully paralyzed but aware) patients. Speech-BCIs aim to map a direct transformation from neural signals to text or speech, which has the potential for a higher communication rate than the current BCIs. Although recent progress has demonstrated the potential of speech-BCIs from either invasive or non-invasive neural signals, the majority of the systems developed so far still assume knowing the onset and offset of the speech utterances within the continuous neural recordings. This lack of real-time voice/speech activity detection (VAD) is a current obstacle for future applications of neural speech decoding wherein BCI users can have a continuous conversation with other speakers. To address this issue, in this study, we attempted to automatically detect the voice/speech activity directly from the neural signals recorded using magnetoencephalography (MEG). First, we classified the whole segments of pre-speech, speech, and post-speech in the neural signals using a support vector machine (SVM). Second, for continuous prediction, we used a long short-term memory-recurrent neural network (LSTM-RNN) to efficiently decode the voice activity at each time point via its sequential pattern-learning mechanism. Experimental results demonstrated the possibility of real-time VAD directly from the non-invasive neural signals with about 88% accuracy.