L2M NSERC - An EMG-based biofeedback system to facilitate neuroplastic adaptation in healthy and clinical populations

运动专业人士，包括物理治疗师和运动教练或训练师，致力于通过识别和解决通常难以视觉检测的运动策略中的细微偏差来修改人类运动。关于肌肉激活策略的有用信息可以使用肌电图（EMG）获得，但这需要专业的信号采集和解释。生物反馈系统已经被开发出来，通过自动化将EMG信号转化为对客户有用的反馈，来增强对EMG信号中包含的有价值信息的访问。然而，这些系统的影响是有限的，因为它们只整合了一两块肌肉的EMG。 姿势调整和其他分布的肌肉活动不被捕获，并且仍然必须由受过训练的锻炼专业人员根据视觉检测通过口头或身体暗示来解决。为了解决这个缺点，我们设计了一个系统，将四个或更多的肌肉源集成到有用的EMG生物反馈中。首先，在特定于训练干预的选定运动的若干重复期间记录EMG信号。这些都是在运动专业人士的监督下进行的，他们提供指导线索，以确保产生最佳的运动模式。该EMG数据用于构建机器学习模型，该模型可以实时识别正在执行的动作。该模型使用户能够控制俄罗斯方块游戏，其中，当相关联的最佳EMG模式已经匹配时，移动对应于块操作。 机器学习模型的预测整合了来自所有传感器的数据，因此可以捕获大部分有助于训练运动的分布式肌肉激活。NSERC Lab2Market计划将加快我们的产品交付给那些可以通过开发评估和最大限度地发挥我们设备商业潜力所需技能而受益的人。我们的系统有可能使生活在影响肌肉控制的条件下的加拿大人受益（例如，中风、脊髓损伤、多发性硬化等），并且还可以帮助训练熟练的运动员或工人，其中改变运动模式可以提高性能或防止受伤。

Exercise professionals, including physiotherapists and athletic coaches or trainers, work to modify human movement by identifying and addressing subtle deviations in movement strategy that are often difficult to detect visually. Useful information about muscle activation strategies can be obtained using electromyography (EMG), but this requires expert signal acquisition and interpretation. Biofeedback systems have been developed to enhance access to the valuable information contained in EMG signal by automating its translation into useful feedback for the client. However, the impact of these systems is limited because they integrate EMG from just one or two muscles. Postural adjustments and other distributed muscle activities are not captured and must still be addressed by verbal or physical cueing pursuant to visual detection by a trained exercise professional. To address this shortcoming, we have designed a system that integrates four or more muscle sources into useful EMG biofeedback. First, EMG signal is recorded during several repetitions of selected movements that are specific to the training intervention. These are performed under the supervision of an exercise professional who provides guiding cues to ensure optimal movement patterns are produced. This EMG data is used to build a machine learning model that can identify movements being performed in real-time. The model enables the user to control a Tetris game wherein movements correspond to block manipulations when the associated optimal EMG pattern has been matched. Predictions by the machine learning model integrate data from all sensors, and can thus capture much of the distributed muscle activations that contribute to trained movements. The NSERC Lab2Market program will accelerate the delivery of our product to those who can benefit by developing skills required to evaluate and maximize the commercial potential of our device. Our system has potential to benefit Canadians living with conditions that impact muscle control (e.g., stroke, spinal cord injury, multiple sclerosis etc.), and can also aid in the training of skilled athletes or workers where modifying movement patterns can improve performance or prevent injury.