Xu Jianbo, Cui Peng, Chen Wenxi
Annu Int Conf IEEE Eng Med Biol Soc. 2020 Jul;2020:5276-5279. doi: 10.1109/EMBC44109.2020.9176107.
This paper proposes a novel identity validation method using ECG signal measured during bathing at 5 different bathtub water temperature ranges, that are 37 ±0.5 °C, 38±0.5 °C, 39±0.5 °C, 40±0.5 °C and 41±0.5 °C, respectively. The experiment includes 5 male and 5 female subjects, each subject collects 2 ECG recordings at each bathtub water temperature range, one day one recording, 10 ECG recordings are collected from each subject, each ECG recording is 18 minutes long, the sampling rate is 200 Hz. During the data processing stage, we perform spectrum analysis, baseline wandering removal, 50 Hz electromagnetic interference removal, signal smoothing, R peaks detection, and QRS complex segmentation. During the classification stage, we perform identity validation using long short-term memory (LSTM) classification network. 5 classification models are trained based on different bathtub water temperature ranges and the cross-validation method is used. Preliminary validation results show that different bathtub water temperature has an important impact on the identity validation. In order to precisely and quickly perform identity validation at different bathtub water temperature ranges, the final classification model is trained based on the samples from 5 different bathtub water temperature ranges. The highest and average identity validation accuracies are 98.43% and 97.68%, respectively.
本文提出了一种新颖的身份验证方法，该方法使用在5个不同的浴缸水温范围内洗澡时测量的心电图信号，这5个水温范围分别为37±0.5°C、38±0.5°C、39±0.5°C、40±0.5°C和41±0.5°C。实验包括5名男性和5名女性受试者，每个受试者在每个浴缸水温范围内收集2份心电图记录，一天收集一份记录，每个受试者共收集10份心电图记录，每份心电图记录时长为18分钟，采样率为200Hz。在数据处理阶段，我们进行频谱分析、去除基线漂移、去除50Hz电磁干扰、信号平滑、R波检测和QRS波群分割。在分类阶段，我们使用长短期记忆(LSTM)分类网络进行身份验证。基于不同的浴缸水温范围训练了5个分类模型，并使用交叉验证方法。初步验证结果表明，不同的浴缸水温对身份验证有重要影响。为了在不同的浴缸水温范围内精确快速地进行身份验证，基于5个不同浴缸水温范围的样本训练了最终的分类模型。最高和平均身份验证准确率分别为98.43%和97.68%。
