Wearable electromyography devices can detect muscular activity for health monitoring and body motion tracking, but this approach is limited by weak and stochastic signals with a low spatial resolution. Alternatively, echomyography can detect muscle movement using ultrasound waves, but typically relies on complex transducer arrays, which are bulky, have high power consumption and can limit user mobility. Here we report a fully integrated wearable echomyography system that consists of a customized single transducer, a wireless circuit for data processing and an on-board battery for power. The system can be attached to the skin and provides accurate long-term wireless monitoring of muscles. To illustrate its capabilities, we use this system to detect the activity of the diaphragm, which allows the recognition of different breathing modes. We also develop a deep learning algorithm to correlate the single-transducer radio-frequency data from forearm muscles with hand gestures to accurately and continuously track 13 hand joints with a mean error of only 7.9°.

可穿戴肌电图设备可以检测肌肉活动以进行健康监测和身体运动跟踪，但这种方法受到空间分辨率低的微弱和随机信号的限制。或者，超声肌电图可以使用超声波检测肌肉运动，但通常依赖于复杂的换能器阵列，这些换能器阵列体积大、功耗高，并且会限制用户的活动能力。在这里，我们报告了一个完全集成的可穿戴超声肌化成像系统，该系统由一个定制的单个传感器、一个用于数据处理的无线电路和一个用于电源的板载电池组成。该系统可以连接到皮肤上，并提供对肌肉的准确长期无线监测。为了说明它的功能，我们使用该系统来检测横膈膜的活动，从而可以识别不同的呼吸模式。我们还开发了一种深度学习算法，将来自前臂肌肉的单传感器射频数据与手势相关联，以准确、连续地跟踪 13 个手关节，平均误差仅为 7.9°。