Non-invasive multi-modal human identification system combining ECG, GSR, and airflow biosignals

A huge amount of data can be collected through a wide variety of sensor technologies. Data mining techniques are often useful for the analysis of gathered data. This paper studies the use of three wearable sensors that monitor the electrocardiogram, airflow, and galvanic skin response of a subject with the purpose of designing an efficient multi-modal human identification system. The proposed system, based on the rotation forest ensemble algorithm, offers a high accuracy (99.6 % true acceptance rate and just 0.1 % false positive rate). For its evaluation, the proposed system was testing against the characteristics commonly demanded in a biometric system, including universality, uniqueness, permanence, and acceptance. Finally, a proof-of-concept implementation of the system is demonstrated on a smartphone and its performance is evaluated in terms of processing speed and power consumption. The identification of a sample is extremely efficient, taking around 200 ms and consuming just a few millijoules. It is thus feasible to use the proposed system on a regular smartphone for user identification.

sensor data; bioinformatics; human identification; data mining; ensemble classification; classifier ensemble method; recognition

Non-invasive multi-modal human identification system combining ECG, GSR, and airflow biosignals Articles