Non-invasive optical sensing of muscle activity.

非侵入性检测肌肉活动的领域标准技术是肌电(EMG)。肌电传感器有许多局限性，包括对电噪声的敏感性，以及传感器必须与肢体表面有稳定的机械耦合。本项目将研究使用一种新的技术来利用红外光来感知肌肉活动的可行性。这项工作的第一阶段将涉及设计由模块化组件组成的光学传感器。初步测试将使用不同的电路元件和设计来收集经验数据。将对光在组织中的传播进行生物物理分析。经验数据和数学模型将确定最佳电路元件和参数。第二阶段将重点放在将多个光学传感器组合成阵列，用于数据采集。参与者将在佩戴阵列和肌电传感器的同时进行抓握和手指移动。简单的线性分类方法将根据肌电数据对光学数据进行基准测试。该项目的第三阶段将侧重于光学数据的实时信号处理，以用作控制信号。这项工作可能集中在光学信号的非平稳性质以及从这些类型的信号中提取有意义的特征的最佳方法。对数据训练、黑盒分类器和由此产生的分类器结构进行分析的黑盒方法将确定哪些光学特征最相关。进一步的分析将包括传感器位移对读数的影响。这对假肢用户特别感兴趣，他们可能会在执行日常活动时发生小碰撞后发现自己的设备不可预测。最终的传感器将是一种电池供电的无线可穿戴设备，能够远程记录数据，并在临床假肢插座中进行测试。

The field-standard technology for non-invasive sensing of muscle activity is Electromyography (EMG). Electromyography sensors have a number of limitations, including susceptibility to electrical noise and a requirement that the sensors have a stable mechanical coupling to the limb surface.This project will study the feasibility of using a novel technique for sensing muscle activity using infrared light. The first stage of the work will involve designing an optical sensor comprised of modular components. Preliminary tests will collect empirical data using varying circuit components and designs. A biophysical analysis of light propagation through tissue will be undertaken. Empirical data and mathematical models will determine optimal circuit components and parameters. The second stage will focus on combining multiple optical sensors to form an array, used for data acquisition. Participants will perform grasps and finger movements while wearing the array and EMG sensors. Simple linear classification methods will benchmark optical data against EMG data. The third phase of the project will focus on real-time signal processing of optical data for use as a control signal. This work is likely to focus on the non-stationary nature of optical signals and the best methods for extracting meaningful features from these types of signals. A black-box approach whereby data trains black box classifiers and the resulting classifier structures are analysed will determine what optical features are most relevant.Further analysis will comprise the impact on readings following a sensor displacement. This is of particular interest for prosthesis users who may find their devices unpredictable after small collisions while performing daily activities. The final sensor will be a battery-powered, wireless wearable device capable of remote datalogging, tested within clinical prosthetic sockets.