Unobtrusive and Continuous BCG-Based Human Identification Using a Microbend Fiber Sensor

Biometric-based human identification has been investigated and applied in daily life applications, but some of them meet fundamental limitations, e.g., highly sensitive to the ambient light of the image-based recognition and unable to detect wrinkle finger with the fingerprint identification, and therefore, decrease the user experience. Recently, electrocardiogram (ECG) has demonstrated to be a very attractive identification technique against attacks. However, the ECG-based human identification is still not a convenient solution since direct skin-contact and chest/body hair shaving are required for ECG acquisition even with the novel single-lead ECG. We propose and experimentally demonstrate an unobtrusive and continuous ballistocardiogram (BCG)-based human identification system using a microbend fiber sensor, which provides a more convenient way to identify the human who is seating on the chair with leaning back against the specially-designed cushion. To the best of our knowledge, this is the first study that a microbend fiber sensor is exploited for human identification. The results show that the proposed 1D-convolutional neural network (CNN) delivers outstanding performance with an average of 100% and 90% identification rate under 15 subjects and 30 subjects, respectively. We recommend that the BCG-based biometric identification technique has great potential to be an innovative solution in smart, private, and security applications with a small group.