Platypus — Indoor Localization and Identification through Sensing Electric Potential Changes in Human Bodies

Outline Introduction Background System overview Model evaluation

Introduction Platypus: 鸭嘴兽 “鸭嘴兽在水中并不依赖于视觉、听觉和嗅觉寻找猎物，而是通过电信号和力学波来探测猎物所在的位置。”

Introduction Platypus: Localize and identify people Without special hardware Remotely and passively sense changes in body electric potential during walking

Background As people move, their bodies’ electric potential changes when interacting with the environment Electric potential of the human body: triboelectric effect capacitive coupling 环境中的任何物体都携带电能，不同物体之间因为携带电能不同而会有电势差，人体的心脏跳动和肌肉运动会带来电能，本文中我不考虑这种内部产生的电能 在人穿着鞋子走路的时候，会有两种主要的方法产生电能 当鞋子和地面接触时，鞋底和地面的材料通过接触而产生电能这就是摩擦生电 在走路的时候，因为鞋底和地面间距离的变化，使得人的身体和地面间的电容耦合变化，当抬脚的时间减小，反之增大。

Background Measure Electric potentials:electric potential sensor The amplitude of a sensor signal increases when the person moves closer to the sensor and vice versa

System overview Rely on electric potential sensors attached high on walls or on ceilings A minimum of three sensors allows us to estimate the position of a person The body electric potential exhibits a characteristic pattern enabling us to distinguish people from each other introducing a model that predicts how sensor values change when a human affects the ambient electric field. inverting the model to derive the position and person 1.主要依赖于在天花板上或者高处的墙上的传感器来获得人体内的电势变化

Model describe a person’s position and change in body electric potential based on the physical principles of the sensing modality Estimate sensor values based on a person’s location and body electric potential EPS: be modelled by a resister and a capacitor，as well as a voltage follower 不同人在走路时的电势变化是有非常明显的不同的，因为不同人的步态以及走路习惯等很不同

Model Feature extraction process 这是一个特征提取过程 尽管步态特征的幅度可能会不同，但是它们的形状大概是一致的 为了获取步态的频率特征，用海明窗和每一个步态特征相乘，并把它们串联起来 图三计算了平均的步态长度和一步持续时间，以及标准偏差

Evaluation electric field exponent α：α= 2.5 performs the best 当α的值为2.5时，错误率最低

Evaluation a standard calibration is sucient for the use of Platypus Slower walking speeds lead to smaller errors 0.16 m for normal walking speed and 0.13 m for slow walking speeds No correlation between shoe size and body weight with regard to the localization error could be observed. The effect of the calibration factor hardly impact the design of the system,一个标准的校验对于P就够了，因为校验因子不会影响系统的设计 更慢的行走速度会让错误更小

Evaluation two participants can be distinguished almost perfectly. The performance degrades with the number of people Only slight differences in classification performance could be observed for walking slowly and normally. 如果是两个人的话，可以完美的进行区分 对着人数的增加而性能变差 分类性能对于走的慢和正常的差别很小很难被观察到

Limitations & summary Platypus operates without the need for person-specific sensor calibration independent of the user’s height/weight/clothing &&the presence or absence of grounded electrical objects person-specific model training is still required identification of users depends on wearing the same footwear. walls have a very strong coupling to users , which could create blind spots in a deployed Platypus system near walls