Wu Yu Tzu, Gomes Matheus K, da Silva Willian Ha, Lazari Pedro M, Fujiwara Eric
Laboratory Photonic Materials and Devices, School of Mechanical Engineering, University of Campinas, Campinas, Brazil.
Biomed Eng Comput Biol. 2020 Mar 24;11:1179597220912825. doi: 10.1177/1179597220912825. eCollection 2020.
Force myography (FMG) is an appealing alternative to traditional electromyography in biomedical applications, mainly due to its simpler signal pattern and immunity to electrical interference. Most FMG sensors, however, send data to a computer for further processing, which reduces the user mobility and, thus, the chances for practical application. In this sense, this work proposes to remodel a typical optical fiber FMG sensor with smaller portable components. Moreover, all data acquisition and processing routines were migrated to a Raspberry Pi 3 Model B microprocessor, ensuring the comfort of use and portability. The sensor was successfully demonstrated for 2 input channels and 9 postures classification with an average precision and accuracy of ~99.5% and ~99.8%, respectively, using a feedforward artificial neural network of 2 hidden layers and a competitive output layer.
在生物医学应用中，力肌动描记法（FMG）是传统肌电图的一种有吸引力的替代方法，主要是因为其信号模式更简单且不受电干扰影响。然而，大多数FMG传感器会将数据发送到计算机进行进一步处理，这降低了用户的移动性，从而减少了实际应用的机会。从这个意义上讲，这项工作提出用更小的便携式组件对典型的光纤FMG传感器进行重新设计。此外，所有数据采集和处理程序都迁移到了树莓派3 B型微处理器上，确保了使用的舒适性和便携性。使用具有2个隐藏层和竞争输出层的前馈人工神经网络，该传感器在2个输入通道和9种姿势分类中得到了成功验证，平均精度和准确率分别约为99.5%和99.8%。
