Neural populations can maintain stable representations of navigation-related variables while integrating uncertain sensory signals. Experimental evidence showed that the precision of head-direction (HD) representations in flies and mice depends on the reliability of sensory cues, highlighting the influence of input uncertainty in attractor-based neural circuits. How do neural dynamics maintain stability while computing under uncertainty? Here, we propose a spiking neural network that unifies two principles - stability through attraction and uncertainty through fluctuation - and reinterpret the HD circuit as an uncertainty-aware integrator rather than a deterministic compass. Specifically, the network uses sampling-based probabilistic inference, where a neural population represents input uncertainty by rapidly fluctuating among likely hypotheses about the world while preserving a stable representation of head direction along an attractor manifold. This formulation suggests why a classical HD "bump" becomes less precise, namely due to rapid fluctuations, reflecting the uncertainty in angular velocity inputs. Our implementation yields experimentally testable predictions: correlated subthreshold voltage fluctuations, multi-timescale nonlinear interaction patterns, and characteristic statistics of bump movement. By combining probabilistic inference with attractor dynamics within one single circuit, our framework suggests how neural populations across species can represent an estimate and its uncertainty through fluctuations while maintaining stability, which could be a general principle for uncertainty-aware computation in noisy biological systems.

Background and Objectives: Robotics training improves gait after stroke, but no prior studies have investigated whether emerging long-term gait biomechanics improvements occur after training. We assessed the temporal profile of pre-post gait biomechanics changes after 9 weeks of dorsiflexion specific adaptive control ankle robot (AMBLE™) training, and at 9 weeks post-training and 17 months later in three persons with chronic stroke to probe for ongoing locomotor plasticity versus post-training disuse decay. Materials and Methods: Three densely hemiparetic subjects (mean ± SD), age 62 ± 7 years., stroke latency 8 ± 4 years, available for repeat testing from an original N = 24 robotics training cohort study, underwent three-dimensional gait analyses pre-post 9 weeks of AMBLE training, and then 9 weeks and 17 months after all robotics training ended. Results: We found that only 47% of total improvements in heel-first strikes and 31% increased paretic step length occurred pre-post training. Unexpectedly, all other biomechanical improvements manifested progressively 17 months after training ended, including ankle peak swing angle (∆ = 7°), dorsiflexion angular velocity (∆ = 23°/s), peak knee flexion (∆ = 11.1°) and hip flexion (∆ = 6°). Robotics prescription progressions in level of assistance and dorsiflexion target angle strongly correlated to gait biomechanical outcomes at 17 months, including improved heel-first strikes and peak dorsiflexion swing angle in this small sample. Conclusions: These findings show that initial improvements in foot-ankle function across training are followed by emergent biomechanical improvements in ankle, knee and hip kinematics across 17 months post-training, with delayed outcomes related to robotics prescription progression. The temporal profile of biomechanical adaptations might suggest delayed, progressive reduction in pathological multi-joint synergies of the hemiparetic leg. However, findings are exploratory and cannot establish causality, treatment efficacy or broad generalizability. Future research is needed to determine whether ankle robotics training can catalyze improvements in long-term gait biomechanical safety and efficiency in the chronic disease management of stroke.

Surface electromyography (sEMG) signals directly reflect movement intention and are therefore promising for natural human-machine interaction. However, their inherent non-stationarity and high inter-subject variability remain major obstacles to robust feature extraction and model generalization. To address these challenges, this study proposes a dual-stream spatial pyramid pooling convolutional neural network (DSSCNN). In this framework, one-dimensional sEMG segments are transformed into two complementary image representations, continuous wavelet transform (CWT) spectrograms and Gramian angular difference field (GADF) images, forming a dual-channel input that jointly preserves time-frequency dynamics and temporal correlation structures. A dual-stream convolutional architecture then extracts discriminative features from each modality, after which a spatial pyramid pooling (SPP) layer aggregates multi-scale representations, enhancing the network's capacity to capture robust spatiotemporal patterns. Extensive experiments demonstrate that DSSCNN achieves an average gesture recognition accuracy of 97.88% with low inter-subject variance under intra-subject random split, and 96.59% under leave-one-subject-out (LOSO) protocol. The practical viability of the proposed approach is further validated through real-time control of an unmanned ground vehicle (UGV). These results not only indicate that the dual-stream framework combined with SPP layer provides an effective strategy for high-precision sEMG-based gesture recognition but also provides a promising technical pathway toward next-generation natural human-machine interaction.

Objectives: Knee proprioception is affected by many biomechanical and physiological factors. Often, during training, rehabilitation, or specific sport requirements, weight or pressure is applied to the foot. However, it is not clear if such applications affect knee proprioceptive acuity. This study examined whether the application of an external weight (3 kg) or of pressure (120 mmHg) at a level above the ankle joint would affect knee proprioception in an open kinematic chain movement in resting or fatigue conditions. Methods: Participants included active young men (n = 7) and women (n = 15), aged 21-34 years, without prior knee injury. Women were tested in their follicular phase. An isokinetic dynamometer was used to evaluate knee joint repositioning. Three knee angles were targeted (30°, 45°, 60°) before and after localized muscle fatigue. A three-way ANOVA analysis with repeated measures and one independent variable (gender) was performed. Results: Analysis showed that ankle weight application positively influenced knee joint proprioceptive acuity resulting in an overall reduction in knee joint angular error for both genders, with the "corrective" effect most evident at 45° and 60° knee joint angles (p < 0.05), whether in the rested or fatigued state. The application of pressure however improved knee proprioceptive performance in men (p < 0.05), in both the rest and fatigue states, but not in women (ns tendency). Conclusions: Application of a small weight at the ankle level significantly improved proprioceptive knee joint acuity. These findings may have wider applicability by allowing the development of specific preventive measures towards safeguarding weaker and/or less resilient players.

Precise coordination between neurons and cortical areas is crucial for sensory perception, yet in vision, continuous fixational eye movements constantly shift retinal input. Traditionally considered random, these drifts modulate neural activity and may challenge the temporal precision needed for perceptual integration, particularly given their partial interocular independence. Recent evidence suggests that ocular drifts are not purely stochastic and play an active role in visual processing. Here, we investigated the impact of fixational eye movements on visual perception using a figure-ground segmentation task. Participants maintained central fixation while viewing textures composed of oriented line segments and reported the presence or absence of a centrally embedded figure. We simultaneously recorded binocular drift and microsaccades, analyzing interocular coordination through 2D cross-correlograms of angular velocities. We found that figure-presents trials exhibited narrow peaks in interocular correlation, while figure-absent trials showed broader, less synchronized drift patterns. Microsaccades were observed to facilitate drift alignment across the eyes. The decrease in microsaccade rate in figure-present trials may contribute to the change in drift synchrony, but it is not required for it. Changes in eye synchrony are also observed under monocular viewing conditions, whereas variations in accommodation and pupil size appear to be unrelated. These results identify two distinct ocular drift modes associated with different perceptual outcomes. Our findings suggest that ocular drift operates under closed-loop control and contributes to effective visual processing.

In environments with multiple talkers, humans can 'tune in' to a speaker of interest while ignoring competing voices. In such conditions, however, auditory cortices track the attended speech envelope rhythms (cortical tracking of speech, CTS) less accurately than in quiet, hindering intelligibility. Visual speech cues (e.g., lip movements) can enhance this CTS, but it remains unclear whether other non-auditory sensory cues, such as tactile input, provide comparable benefits through similar neural mechanisms. Here, using magnetoencephalography, we quantified CTS as the coherence between the speech temporal envelope of the attended speaker and brain responses in supratemporal auditory areas at syllabic (4-8 Hz), word (1-4 Hz) and phrasal/sentential (<1 Hz) frequencies. Participants listened to connected speech presented alone, together with synchronous or asynchronous speech-based vibrations, or with the corresponding speaker video, in both quiet and a multi-talker background. We hypothesized that, in the presence of competing background speakers, speech-based vibrotactile stimulation improves comprehension by enhancing CTS of the attended speaker and modulating auditory-seeded functional brain connectivity with extra-auditory neocortical brain areas. Results revealed that synchronous vibrotactile stimulation improved comprehension in the multi-talker background and increased syllabic CTS at the right auditory cortex, with this CTS increase magnitude correlating with comprehension performance. Audio-tactile CTS enhancement was accompanied by stronger beta-band auditory cortex connectivity with ipsilateral angular and ventral inferior temporal gyri, alongside reduced alpha-band coupling with the precuneus. These findings suggest that vibrotactile input can support speech-in-noise processing by impacting both local auditory cortical activity and auditory-seeded long-range functional connectivity.

The integration of high-density surface electromyography with ultrasound-transparent electrodes (HDEMG-US) and B-mode ultrasonography enables concurrent assessment of motor unit (MU) firing properties and fascicle length (FL) from the same muscle region. Although the isometric relations between MU discharge, FL, and torque have been characterised, their interactions during anisometric contractions remain poorly understood. This study examined MU discharge behaviour and its relation with FL changes during shortening and lengthening contractions of the tibialis anterior muscle. Ten healthy males performed isokinetic dorsiflexion contractions at a constant angular velocity of 2°/s while maintaining a constant submaximal torque of 25% maximum voluntary contraction torque, measured at a long muscle-tendon length (~30° of plantarflexion), with HDEMG-US signals and ultrasound images recorded simultaneously. MUs were identified using blind-source separation decomposition and FL was tracked with an optical flow algorithm. Cross-correlation analysis revealed high associations between cumulative spike train (CST), FL, and torque in both contraction phases. Neuromechanical delays were significantly longer during lengthening than shortening (phase effect: F(1, 7) = 8.51, P = 0.023; mean difference: -25.04 ms, 95% CI: -45.35 to -4.74). Discharge rate changed at a similar rate with variations in FL during both shortening and lengthening contractions (similar discharge rate-FL slopes, P = 0.44); however, absolute discharge rates were systematically lower during lengthening (difference in intercepts: P < 0.0001). These findings demonstrate the feasibility of HDEMG-US for characterising neuromechanical coupling during dynamic contractions and provide new evidence of neural drive regulation during shortening and lengthening muscle actions.