Peking Union Medical College Hospital Licenses $600K Rehabilitation Patent for Human Motion Intention Prediction to Beijing Zhuzheng Robotics

Recently, Peking Union Medical College Hospital (PUMCH) of the Chinese Academy of Medical Sciences publicly announced its plan to transfer the patent rights for the “Method and Device for Predicting Human Motion Intent During Rehabilitation Training” to Beijing Zhuozheng Robotics Co., Ltd. at a proposed transaction price of RMB 600,000.

Professor Zhao Yu, the principal inventor of this patent, is a Professor of Orthopedics at Peking Union Medical College Hospital, Chinese Academy of Medical Sciences. Professor Zhao has conducted extensive research on conditions such as spinal deformities and thoracic spinal stenosis, utilizing genomics, metabolomics, biomechanics, human factors engineering, and artificial intelligence. He has published more than 50 papers in renowned international orthopedic journals, including JBJS (Am), European Spine Journal, and Spine, with over 20 of these articles indexed by SCI.

The other party to the transaction, Beijing Zhuzheng Robotics Co., Ltd., is an innovative enterprise centered on surgical robots, dedicated to providing comprehensive digital solutions for spinal surgery procedures, and building China’s strongest portfolio of medical robot products featuring navigation, positioning, and autonomous operation.

With the intensifying aging of the population, there is a growing incidence of cognitive and balance impairments in patients caused by conditions such as stroke and hemiplegia. Traditional rehabilitation methods are facing challenges due to low efficiency and a shortage of professional personnel. To address this issue, robot-assisted rehabilitation devices, such as balance rehabilitation platforms and exoskeleton robots, have been developed. These devices utilize force sensors to detect patients' movement intentions, thereby enabling human-robot interactive training. However, signal latency in force sensors and time delays in the control systems of training platforms cause the platform's movements to lag behind the patients' intended motions, which adversely affects the efficacy of the training.

Based on this, Professor Zhao Yu’s team has developed a method and device for predicting human motion intent during rehabilitation training, which eliminates the delay error between the movement of the rehabilitation training platform and the user’s motion intent, thereby improving synchronization and enhancing training outcomes.

The core of this method lies in the application of deep learning network models. First, the team collects plantar pressure data and joint posture data from the human body, inputs these data as time series into a pre-trained deep learning network model, and the model can accurately predict human motion intent. This model consists of four major components: a data patch embedding module, an encoder layer, a first time-series decomposition layer, and a decoder layer.

Among these components, the data patch embedding module serves as the data preprocessing stage, responsible for segmenting the raw input time series into patches of a predetermined length and applying positional and value encoding to lay the foundation for subsequent processing. The encoder layer focuses on extracting the first-period features of the data patches, providing critical information for subsequent feature fusion.

Subsequently, the first time series decomposition layer decomposes the raw input time series into initial periodic features and initial trend features, providing more refined data support for subsequent feature computation. The decoder layer then computes the aggregate trend features and aggregate periodic features based on these inputs, and sums them to obtain the final predicted sequence.

Furthermore, to further enhance prediction accuracy, the team introduced a preprocessing module. This module performs stationarization on the raw input time series, eliminating non-stationary components to ensure the stability and reliability of the input data.

It is worth mentioning that,During the experimental phase, the collected pose sequence data were scientifically partitioned into training, validation, and test sets in a 6:2:2 ratio. The results demonstrated that the novel network model achieved superior prediction accuracy and transmitted predicted values to the control software with reduced latency, significantly enhancing the synchronization between the rehabilitation training platform and patient posture, thereby maximizing the platform’s therapeutic efficacy.

Rehabilitation robots, as an innovative product integrating advanced technology with medical needs, have garnered significant attention. In the Chinese market, as the population ages and the number of patients with chronic diseases continues to rise, demand for rehabilitation medical services is growing vigorously. Meanwhile, national policy support provides a strong guarantee for the development of the rehabilitation robot industry.

In recent years, the development of rehabilitation robots in China has also undergone a qualitative transformation.Whether rehabilitation or assistive products, greater emphasis is being placed on intelligent and comfort-oriented development. Not only are the applicable body parts, functions, and training modes becoming increasingly diverse, but comfort levels have also seen significant improvement.Meanwhile, these products are continuously exploring and expanding their application scenarios to meet the needs of more patients.

Taking Anjie Lai Technology as an example, its LiteStepper® single-lower-limb hemiplegia rehabilitation robot and ExoMax® dual-ankle rehabilitation training robot are based on the clinical theory of closed-loop neurorehabilitation grounded in brain plasticity, providing patients with full-cycle, active, efficient, and personalized rehabilitation solutions.

For another example, Siyi Intelligence has launched a hand function rehabilitation robot based on brain-computer interface (BCI) technology. Its EEG cap collects and records electroencephalogram signals in real time, decodes motor intent through intelligent algorithms, and converts it into motion commands to assist movement of the affected hand, thereby achieving mind-controlled operation. Through BCI-based rehabilitation training, it stimulates patients’ voluntary motor awareness and realizes a bidirectional closed-loop neural stimulation of “perception-control.”

In addition, Chinese enterprises have also launched numerous cutting-edge products for home-based scenarios.These products not only meet the demands for diverse event venues and extensive operational ranges, but also reach new heights in safety and technological integration. Meanwhile, companies are continuously enhancing their soft power to address the personalized needs and cost-effectiveness pursuits of different consumer groups.

For patients with paraplegia, Bangbang Robotics has launched an intelligent standing wheelchair. Meanwhile, its intelligent assistive mobility robot helps individuals with mobility impairments achieve independent living at home, and its assistive travel robot enables users with limited mobility to travel safely and stably. These innovations meet the expectations of those facing mobility challenges for "intelligent services + proactive safety protection + multi-environment adaptability"—a suite of warm, cutting-edge technologies.

With continuous technological advancements and ongoing market expansion, the rehabilitation robotics industry is poised for even broader development prospects. We have every reason to believe that, in the near future, rehabilitation robots will become invaluable assistants on the path to recovery for more patients.