China

Six Fujian government departments jointly issued a provincial BCI action plan for 2026–2030. It calls for key technical breakthroughs, more than five innovative small and medium-sized companies, one cross-strait joint laboratory and two provincial innovation platforms by 2027; by 2030, Fujian aims to have 30 companies, a complete industry chain and an industry worth more than 3 billion yuan.

A systematic review of BCI medical applications published on August 17, 2026 in Theoretical and Natural Science covers literature review, comparative case analysis of invasive vs non-invasive techniques, and interdisciplinary assessment. It reports that invasive BCIs reach 80% to 100% task success rates in robotic arm control but are limited by surgical hazards, progressive signal deterioration and costs above $250,000, while non-invasive BCIs are safer and more widely deployed in community neurorehabilitation with about 70% effectiveness for post-stroke upper-limb recovery, yet suffer poor signal-to-noise ratios, a BCI illiteracy rate near 30% and low information transfer speeds. The review also addresses neural data privacy, autonomy paradoxes and inequitable access.

Researchers propose MRieHy, a multi-feature Riemannian hypergraph framework for online test-time adaptation of motor imagery BCI decoding. It aligns multi-day distributions via Riemannian means of covariance matrices, builds one hypergraph with Riemannian distance and a second with cosine similarity, fuses them with adaptively learned weights, and decodes buffered online samples after Riemannian alignment. On a private four-class ECoG dataset and two public four-class EEG datasets, MRieHy shows notable gains over state-of-the-art baselines, targeting the cross-day transferability and online operation that clinical MI-BCI still lacks.

Eleven Chinese provincial-level jurisdictions had published prices for brain-computer interface medical services. The invasive implantation fee runs from RMB 4,389 at the lowest hospital tier in Sichuan to RMB 7,980 in Beijing (approx. $650 to $1,180 at RMB 6.74 to the dollar, the mid-August 2026 rate), and the non-invasive fitting fee from RMB 643 to 990. Sichuan, Hainan, Qinghai and Xinjiang tier prices by hospital grade, Shanghai leaves the amount to each hospital, and Jiangsu's three figures appear only in a July 2025 draft. Beijing lists all three fees as Class A at four named hospitals; Zhejiang covers the fitting fee as Class B, limited to prosthetic fitting. Sichuan, Jiangxi, Hainan, Qinghai and Xinjiang assign Class C, Shanghai excludes the fees, and Hubei and Guangdong have prices without a coverage decision.

The study was published in ITM Web of Conferences on August 14, 2026, addressing the demand for portable, real-time brain-computer interface systems in stroke rehabilitation by completing the physical integration and online experimental validation of a wearable system. The system uses a specialized EEG headset with miniaturized acquisition circuits secured via pogo pins, featuring 10 core recording channels positioned over the sensorimotor cortex. During the evaluation phase, the research team recruited 6 healthy subjects and 2 stroke-affected hemiplegic patients for closed-loop experiments based on motor imagery and motor attempts. Common Spatial Pattern was used for spatial feature extraction and Linear Discriminant Analysis for intention classification, with personalized sub-band optimization applied to further improve recognition. The authors report an average offline recognition rate of 84.91% and a classification accuracy of 79.38% in the more challenging online real-time testing. Analysis of spatiotemporal spectra and R² value distributions validated activation patterns in the sensorimotor areas during motor intention triggering, which the authors present as support for advancing the technology from laboratory settings toward community rehabilitation.

WABO appointed Jianwu Dang to its Scientific Advisory Board and named him joint scientist. Dang, a researcher at the Shenzhen Institutes of Advanced Technology and a distinguished professor at Shenzhen University of Advanced Technology, will support task-conditioned intent models that combine EEG, EMG, speech, text and task context; WABO said the work is not aimed at context-free mind reading.

China's brain-computer interface sector recorded 47 funding rounds in the first half of 2026, 38 of them disclosing terms for a combined RMB 5.38 billion (approx. $750 million). A broader tally used by Chinese financial media puts the total at 64 deals and roughly RMB 7.25 billion (approx. $1 billion). Either figure is more than double the whole of 2025 (49 deals, RMB 2.72 billion) at a comparable deal count. The record lists every deal by month — company, technical approach, round, amount, investors — and identifies three structural patterns: ultrasound and non-invasive approaches moving quickly into a field long defined by implants, Big Tech and municipal state funds entering at the same time, and nearly half of all transactions still sitting at angel or seed stage.

A prospective cohort of 1,641 motor dysfunction patients, 66.24% chronic stroke, provides the first clinical profile for personalised BCI system design.

The General Office of China's Zhejiang Provincial People's Government issued Several Measures for Promoting Industry-University-Research Coordination in Brain-Computer Interfaces, document number Zhe Zheng Ban Fa [2026] No. 33, dated July 16, 2026 and released on August 12, comprising 18 measures in five areas. Class II BCI devices of clear clinical value may enter a special review procedure with the timeline cut to within 40 working days, while medical consumables are listed through a green channel within 10 working days. Funding comes in four tiers: up to 10 million yuan for a provincial major science and technology project, up to 8 million yuan for research hospitals deeply involved in preclinical work and product launch, up to 3 million yuan for applied basic research, and 1 million, 500,000 and 100,000 yuan for leading international, national and local standards. Beds used solely for clinical research are exempt from efficiency and DRG performance assessment.

A completed RCT compared BCI-controlled versus conventional exoskeleton gait training in stroke patients, training 2 times a week for 4 weeks.

A review systematically examines stratified BCI rehabilitation strategies and comprehensive nursing models for paraplegic patients, showing BCIs can activate neuroplasticity and relieve spasticity.

A review analyses technical, hardware, and user-level bottlenecks facing SSVEP-BCI systems in real-world deployment, summarising solutions including EMG fusion denoising, novel sponge electrodes, and hybrid paradigms.

Researchers propose CLMT-Net, a contrastive learning network based on a multi-scale Transformer, for fine-grained within-limb MI decoding, achieving 76.13% accuracy on the MI-2 dataset.

A game theory-based adaptive human-machine joint learning method for online MI-BCI decoding improves accuracy by 9.7% and 5.6% in 14-subject online experiments (both p < 0.01).

Brain-computer interface decoders can guess the right label using information unrelated to the target mental state. A team at the School of Psychology, Beijing Language and Culture University (BLCU), proposes a falsifiable substitution-test standard: candidate evidence must persist in disjoint data, survive capacity-matched substitutions of physical organization or listener templates, and remain testable after target events are excluded. Across six EEG datasets (41 participants), averaging four neural-speech margin metrics brought 5-second decoding to what the authors call a leading level; in two hierarchical interfaces, parent-stream error scores kept AUCs of 0.968 and 0.965 after all target-command events were excluded. The framework offers a test for attributing evidence in neuroscience and BCI.

WABO (瓦博科技) says it ran a four-day neural art experience at the Snapdragon Pavilion during ChinaJoy 2026, held from July 31, 2026 to August 3, 2026 in Shanghai, producing more than 100 AI artworks with participants. Visitors wore a non-invasive EEG headband and viewed feedback derived from their signal state, with task-specific features feeding the artwork-generation process; the company says no private thoughts were read. The setup paired WABO's acquisition hardware with a Snapdragon X Elite laptop, keeping signal processing, model inference and image generation on the endpoint, which the company describes as a practical test of a neural-intent workflow outside a controlled development setting.

Researchers propose ReCIL, a rehearsal-based class incremental learning method for cross-subject motor imagery classification that lets a model learn new MI classes sequentially without retraining from scratch. Using Euclidean alignment to reduce cross-subject EEG distribution shift and global-local replay to preserve earlier-task knowledge, ReCIL achieved a good balance between plasticity and stability across three public MI datasets. The authors report it as the first study of cross-subject class incremental learning for MI classification.

Beijing's Haidian district is accelerating its push to become a hub for brain-computer interface development, with nearly 60% of China's key BCI companies based in the district, CCTV.com reported on August 6, 2026. According to the report, the Zhongguancun (Haidian) BCI Industry Cluster was formally unveiled at the 2026 Zhongguancun Forum annual meeting and now hosts more than 20 BCI projects, and an AI-plus-BCI standards testing and validation laboratory has been completed. The report did not disclose a list of the cluster's projects, the methodology behind the company figure, or further construction plans.

Researchers at Air Force Medical University have published a functional near-infrared spectroscopy (fNIRS) dataset in Scientific Data: frontal hemoglobin responses from 55 young adults, each recorded across three color-word Stroop sessions spread over about two weeks, with more than 30 trials per condition. The authors offer it for work on conflict inhibition, for building decoders for neurofeedback training, and for training large-scale cross-subject fNIRS models.

The study proposes BiGSTF-Net, a multimodal architecture that exploits the complementary properties of EEG and functional near-infrared spectroscopy (fNIRS) for cognitive-state decoding: heterogeneous spatio-temporal extractors capture each modality's representations, a modal residual interaction unit provides bidirectional cross-modal guidance, and a spatio-temporal gating unit fuses intra-modal features. Under cross-session evaluation on multiple BCI datasets, BiGSTF-Net consistently outperformed representative multimodal fusion baselines; ablations validated each component and visualizations matched the known neurophysiological features of the two signals.

The study presents AutoMI, a framework that uses LLM multi-agents to automatically and rapidly iterate on motor imagery EEG classification models, combining a Q-learning policy with deterministic rules and integrating planning, execution and output agents with predefined tools, plus experience tracking and rollback. Models built by AutoMI reached 77.62%, 78.08% and 83.02% accuracy on the IV2a, OpenBMI and ECUST-MI datasets — up 18.42%, 9.27% and 19.25% over automated optimization algorithms.

Researchers report FRED, a principled EEG-fNIRS fusion framework for imagined handwriting decoding, posted to arXiv as a preprint on August 4, 2026 and not yet peer reviewed. Imagined handwriting offers a temporally rich paradigm for non-invasive neural decoding, and EEG-fNIRS fusion promises complementary neural information, but fusion is typically heuristic and lacks principled treatment of frequency-band redundancy. FRED builds frequency-decorrelated temporal ensembles for imagined handwriting decoding. The ensemble reaches 0.8076/0.7242/0.7492 accuracy on the public/private/overall test partitions without test-set adaptation or output constraints, and the complete system reaches 0.8498/0.7718/0.7952, ranking fourth on the private split. A modality audit finds fNIRS-only decoding at chance (0.2511 overall), while adding fNIRS to EEG changes accuracy by only +0.0025.

The study recorded hippocampal CA1 population activity in rats performing a sequential spatial decision task in a modified T-maze, dividing the decision into initiation, running and approach phases. Local theta sequences consistently over-represented the actual choice, while prospective representations driven by choice-arm place cells shifted from predicting the actual choice during running to representing potential paths more evenly at the choice point. Integrating local and prospective features improved decoding, reaching 74.4% accuracy for future choice prediction and 78.2% for upcoming trajectory decoding.

Researchers report SAGE, a switch-aware EEG-guided soft gating framework for target speaker extraction, posted to arXiv as a preprint on August 3, 2026 and not yet peer reviewed. Under in-trial auditory attention switching, neural noise and intrinsic latency can delay or destabilize attention tracking, and conventional methods often cause discontinuities at switching points. SAGE treats in-trial switching as dynamic selection, generating two candidate speech streams with a robust separator and using an EEG-guided switch-aware gating module to produce smooth fusion weights and suppress transition artifacts. It integrates latency-compensated alignment and an uncertainty-driven conservative strategy, outperforms baselines, achieves 8.67 dB SI-SDR and 88.24% STOI, and reduces average switching latency to 2.04 s.

People's Daily Online reported that multiple people with high-level paralysis used StairMed's minimally invasive implantable BCI for three to four consecutive hours of computer work, including e-commerce logistics coordination and vending-machine data labeling. The company reported a 256-channel system and latency below 50 milliseconds. The article did not provide a participant count, standardized protocol or adverse-event table.